hatespeech_detection.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "gpuType": "T4"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "source": [
        "#1. Introduction"
      ],
      "metadata": {
        "id": "iz3R9ZK89gYg"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cfU-BBPo9oE_",
        "outputId": "80bc7937-ecf1-4225-af4a-ae62cf891b8b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# 2. EDA"
      ],
      "metadata": {
        "id": "zFlpaRRe9wSm"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Install and import necessary packages"
      ],
      "metadata": {
        "id": "CDkJV4xQCEwP"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!pip install pyvi\n",
        "!pip install emoji"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "FLEQXJEUBzTL",
        "outputId": "cb04a69c-d791-437f-f5fc-465f5860e154"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Collecting pyvi\n",
            "  Downloading pyvi-0.1.1-py2.py3-none-any.whl (8.5 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m8.5/8.5 MB\u001b[0m \u001b[31m28.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: scikit-learn in /usr/local/lib/python3.10/dist-packages (from pyvi) (1.2.2)\n",
            "Collecting sklearn-crfsuite (from pyvi)\n",
            "  Downloading sklearn_crfsuite-0.3.6-py2.py3-none-any.whl (12 kB)\n",
            "Requirement already satisfied: numpy>=1.17.3 in /usr/local/lib/python3.10/dist-packages (from scikit-learn->pyvi) (1.25.2)\n",
            "Requirement already satisfied: scipy>=1.3.2 in /usr/local/lib/python3.10/dist-packages (from scikit-learn->pyvi) (1.11.4)\n",
            "Requirement already satisfied: joblib>=1.1.1 in /usr/local/lib/python3.10/dist-packages (from scikit-learn->pyvi) (1.3.2)\n",
            "Requirement already satisfied: threadpoolctl>=2.0.0 in /usr/local/lib/python3.10/dist-packages (from scikit-learn->pyvi) (3.4.0)\n",
            "Collecting python-crfsuite>=0.8.3 (from sklearn-crfsuite->pyvi)\n",
            "  Downloading python_crfsuite-0.9.10-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.1 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m1.1/1.1 MB\u001b[0m \u001b[31m68.0 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: six in /usr/local/lib/python3.10/dist-packages (from sklearn-crfsuite->pyvi) (1.16.0)\n",
            "Requirement already satisfied: tabulate in /usr/local/lib/python3.10/dist-packages (from sklearn-crfsuite->pyvi) (0.9.0)\n",
            "Requirement already satisfied: tqdm>=2.0 in /usr/local/lib/python3.10/dist-packages (from sklearn-crfsuite->pyvi) (4.66.2)\n",
            "Installing collected packages: python-crfsuite, sklearn-crfsuite, pyvi\n",
            "Successfully installed python-crfsuite-0.9.10 pyvi-0.1.1 sklearn-crfsuite-0.3.6\n",
            "Collecting emoji\n",
            "  Downloading emoji-2.11.0-py2.py3-none-any.whl (433 kB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m433.8/433.8 kB\u001b[0m \u001b[31m3.5 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hInstalling collected packages: emoji\n",
            "Successfully installed emoji-2.11.0\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import os\n",
        "import pandas as pd\n",
        "from sklearn.model_selection import train_test_split\n",
        "import matplotlib.pyplot as plt\n",
        "import seaborn as sn\n",
        "import numpy as np\n",
        "import re\n",
        "import emoji\n",
        "from wordcloud import WordCloud\n",
        "from pyvi.ViTokenizer import ViTokenizer\n",
        "from keras.preprocessing import text, sequence\n",
        "from keras.utils import to_categorical\n",
        "from sklearn.metrics import f1_score, confusion_matrix, accuracy_score\n",
        "from keras.layers import Input, Dense, Embedding, Conv2D, MaxPool2D\n",
        "from keras.layers import Reshape, Flatten, Dropout, Concatenate\n",
        "from keras.callbacks import ModelCheckpoint\n",
        "from keras.optimizers import Adam\n",
        "from keras.models import Model\n",
        "DATA_PATH = \"/content/drive/MyDrive/midtermdata/\"\n",
        "\n",
        "os.listdir(DATA_PATH)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "o7a5Jyk8x4aQ",
        "outputId": "25c4e606-20a5-4c65-ce03-bcf73f236b61"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['cc.vi.300.vec',\n",
              " 'dev.csv',\n",
              " 'test.csv',\n",
              " 'train.csv',\n",
              " 'facebook_comment_2k7.csv',\n",
              " 'facebook_comment_2k7.gsheet',\n",
              " 'train.gsheet',\n",
              " 'vietnamese-stopwords-dash.txt',\n",
              " 'VnCoreNLP-1.1.1.jar',\n",
              " 'Text_CNN_model_v1.h5',\n",
              " 'GRU_model_v1.h5',\n",
              " 'transformer_model']"
            ]
          },
          "metadata": {},
          "execution_count": 4
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Load data 2k7\n",
        "\n",
        "Mục đích xây dựng model phát hiện 3 nhãn riêng biệt: harrassment, hatespeech, normal nên các nhãn chỉ cần có chứa harrassment sẽ là harrassment, có hatespeech sẽ được xem là hatespeech, còn lại là normal.\n",
        "\n",
        "Trong vihsd: CLEAN (0), OFFENSIVE (1), and HATE (2)\n",
        "\n",
        "Trong tập 2k7 hiện tại: harassment = HATE, hatespeech = OFFENSIVE, normal = CLEAN nên nhãn cho tập dữ liệu kết hợp là:\n",
        "'normal': 0, 'hate_speech': 1, 'harassment': 2"
      ],
      "metadata": {
        "id": "ljntnATdCRH0"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Load the dataset\n",
        "FILE_PATH = \"facebook_comment_2k7.csv\"\n",
        "df2k7 = pd.read_csv(DATA_PATH + FILE_PATH)\n",
        "df2k7 = df2k7.drop(columns=['id', 'Unnamed: 0'], errors='ignore')\n",
        "def adjust_label(label):\n",
        "    if 'harassment' in label:\n",
        "        return 2\n",
        "    elif 'hate_speech' in label:\n",
        "        return 1\n",
        "    else:\n",
        "        return 0\n",
        "\n",
        "df2k7['label'] = df2k7['label'].apply(adjust_label)"
      ],
      "metadata": {
        "id": "WpvTVhXN7mYM"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "df2k7['label'].unique()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "MhJTBctsGhi7",
        "outputId": "10486d7f-fd21-4a1f-cfbf-8a90994c9695"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([2, 0, 1])"
            ]
          },
          "metadata": {},
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "len(df2k7)"
      ],
      "metadata": {
        "id": "vHwzs9ocGnTL"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "source": [
        "# @title Label Frequency\n",
        "\n",
        "df2k7['label'].value_counts().plot(kind='bar')"
      ],
      "cell_type": "code",
      "execution_count": null,
      "outputs": [],
      "metadata": {
        "id": "p1O4YlpjTRiA"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Do tính phức tạp và tinh vi của lời nói căm thù và hành vi quấy rối, tập dữ liệu với 2.700 nhận xét có thể không cung cấp đủ ví dụ để mô hình học máy tìm hiểu và khái quát hóa một cách hiệu quả, đặc biệt nếu những nhận xét này không được phân bố đồng đều giữa các class, như hình trên dữ liệu không cân bằng và có số lượng rất ít. Điều quan trọng là mô hình phải tiếp xúc với nhiều kiểu ngôn ngữ, sắc thái và ngữ cảnh khác nhau để hoạt động tốt trên dữ liệu trong thế giới thực và một tập dữ liệu hạn chế có thể không nắm bắt được toàn bộ cách ngôn từ kích động thù địch và hành vi quấy rối có thể biểu hiện trong ngôn ngữ.\n",
        "\n",
        "Do đó, việc đào tạo một mô hình trên tập dữ liệu này có thể dẫn đến một hệ thống không đủ mạnh để xử lý các biểu hiện đa dạng của lời nói căm thù và quấy rối trong các tình huống khác nhau, có khả năng dẫn đến tỷ lệ dương tính giả hoặc âm tính giả cao khi được triển khai trong môi trường trực tiếp.\n",
        "\n",
        "Để giảm thiểu điều này, hãy xem xét các kỹ thuật như tăng cường dữ liệu để mở rộng tổng hợp các lớp thiểu số, sử dụng các mô hình được đào tạo trước hoặc kết hợp các bộ dữ liệu bên ngoài. Cách tiếp cận này có thể mang lại cho mô hình trải nghiệm học tập phong phú hơn và cải thiện hiệu suất của mô hình trong việc phát hiện lời nói căm thù và hành vi quấy rối."
      ],
      "metadata": {
        "id": "90vq_S_sUUoy"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Kết hợp tập dữ liệu ViHSD"
      ],
      "metadata": {
        "id": "XcGeXJUjVYtd"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Chia tập 2k7 thành train, val, test với tỉ lệ 7:1:2"
      ],
      "metadata": {
        "id": "ztjIOSGqnNeh"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Split the data into train+val and test sets with stratification\n",
        "train_val, test = train_test_split(df2k7, test_size=0.2, stratify=df2k7['label'], random_state=42)\n",
        "train, val = train_test_split(train_val, test_size=0.125, stratify=train_val['label'], random_state=42)"
      ],
      "metadata": {
        "id": "HPPPLNKtnlXU"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Load the VIHSD dataset\n",
        "vihsd_train_path = DATA_PATH + 'train.csv'\n",
        "vihsd_dev_path = DATA_PATH + 'dev.csv'\n",
        "vihsd_test_path = DATA_PATH + 'test.csv'\n",
        "\n",
        "vihsd_train = pd.read_csv(vihsd_train_path)\n",
        "vihsd_dev = pd.read_csv(vihsd_dev_path)\n",
        "vihsd_test = pd.read_csv(vihsd_test_path)"
      ],
      "metadata": {
        "id": "nW0YSfL-nt7Y"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "vihsd_train.head()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 206
        },
        "id": "clKV_7R2oaep",
        "outputId": "43a2002d-d38c-4e9c-dd43-c7ad82607027"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                           free_text  label_id\n",
              "0  Em được làm fan cứng luôn rồi nè ❤️ reaction q...         0\n",
              "1  Đúng là bọn mắt híp lò xo thụt :))) bên việt n...         2\n",
              "2           Đậu Văn Cường giờ giống thằng sida hơn à         0\n",
              "3  CÔN ĐỒ CỤC SÚC VÔ NHÂN TÍNH ĐỀ NGHI VN. NHÀ NƯ...         2\n",
              "4  Từ lý thuyết đến thực hành là cả 1 câu chuyện ...         0"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-c5620345-bb33-433c-a029-73a4b9cf6277\" class=\"colab-df-container\">\n",
              "    <div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>free_text</th>\n",
              "      <th>label_id</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Em được làm fan cứng luôn rồi nè ❤️ reaction q...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>Đúng là bọn mắt híp lò xo thụt :))) bên việt n...</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>Đậu Văn Cường giờ giống thằng sida hơn à</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>CÔN ĐỒ CỤC SÚC VÔ NHÂN TÍNH ĐỀ NGHI VN. NHÀ NƯ...</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>Từ lý thuyết đến thực hành là cả 1 câu chuyện ...</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "    <div class=\"colab-df-buttons\">\n",
              "\n",
              "  <div class=\"colab-df-container\">\n",
              "    <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-c5620345-bb33-433c-a029-73a4b9cf6277')\"\n",
              "            title=\"Convert this dataframe to an interactive table.\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\" viewBox=\"0 -960 960 960\">\n",
              "    <path d=\"M120-120v-720h720v720H120Zm60-500h600v-160H180v160Zm220 220h160v-160H400v160Zm0 220h160v-160H400v160ZM180-400h160v-160H180v160Zm440 0h160v-160H620v160ZM180-180h160v-160H180v160Zm440 0h160v-160H620v160Z\"/>\n",
              "  </svg>\n",
              "    </button>\n",
              "\n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    .colab-df-buttons div {\n",
              "      margin-bottom: 4px;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "    <script>\n",
              "      const buttonEl =\n",
              "        document.querySelector('#df-c5620345-bb33-433c-a029-73a4b9cf6277 button.colab-df-convert');\n",
              "      buttonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "      async function convertToInteractive(key) {\n",
              "        const element = document.querySelector('#df-c5620345-bb33-433c-a029-73a4b9cf6277');\n",
              "        const dataTable =\n",
              "          await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                    [key], {});\n",
              "        if (!dataTable) return;\n",
              "\n",
              "        const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "          '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "          + ' to learn more about interactive tables.';\n",
              "        element.innerHTML = '';\n",
              "        dataTable['output_type'] = 'display_data';\n",
              "        await google.colab.output.renderOutput(dataTable, element);\n",
              "        const docLink = document.createElement('div');\n",
              "        docLink.innerHTML = docLinkHtml;\n",
              "        element.appendChild(docLink);\n",
              "      }\n",
              "    </script>\n",
              "  </div>\n",
              "\n",
              "\n",
              "<div id=\"df-be61f96e-b1fa-40f7-af51-cc6005b4c37b\">\n",
              "  <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-be61f96e-b1fa-40f7-af51-cc6005b4c37b')\"\n",
              "            title=\"Suggest charts\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "<svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "     width=\"24px\">\n",
              "    <g>\n",
              "        <path d=\"M19 3H5c-1.1 0-2 .9-2 2v14c0 1.1.9 2 2 2h14c1.1 0 2-.9 2-2V5c0-1.1-.9-2-2-2zM9 17H7v-7h2v7zm4 0h-2V7h2v10zm4 0h-2v-4h2v4z\"/>\n",
              "    </g>\n",
              "</svg>\n",
              "  </button>\n",
              "\n",
              "<style>\n",
              "  .colab-df-quickchart {\n",
              "      --bg-color: #E8F0FE;\n",
              "      --fill-color: #1967D2;\n",
              "      --hover-bg-color: #E2EBFA;\n",
              "      --hover-fill-color: #174EA6;\n",
              "      --disabled-fill-color: #AAA;\n",
              "      --disabled-bg-color: #DDD;\n",
              "  }\n",
              "\n",
              "  [theme=dark] .colab-df-quickchart {\n",
              "      --bg-color: #3B4455;\n",
              "      --fill-color: #D2E3FC;\n",
              "      --hover-bg-color: #434B5C;\n",
              "      --hover-fill-color: #FFFFFF;\n",
              "      --disabled-bg-color: #3B4455;\n",
              "      --disabled-fill-color: #666;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart {\n",
              "    background-color: var(--bg-color);\n",
              "    border: none;\n",
              "    border-radius: 50%;\n",
              "    cursor: pointer;\n",
              "    display: none;\n",
              "    fill: var(--fill-color);\n",
              "    height: 32px;\n",
              "    padding: 0;\n",
              "    width: 32px;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart:hover {\n",
              "    background-color: var(--hover-bg-color);\n",
              "    box-shadow: 0 1px 2px rgba(60, 64, 67, 0.3), 0 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "    fill: var(--button-hover-fill-color);\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart-complete:disabled,\n",
              "  .colab-df-quickchart-complete:disabled:hover {\n",
              "    background-color: var(--disabled-bg-color);\n",
              "    fill: var(--disabled-fill-color);\n",
              "    box-shadow: none;\n",
              "  }\n",
              "\n",
              "  .colab-df-spinner {\n",
              "    border: 2px solid var(--fill-color);\n",
              "    border-color: transparent;\n",
              "    border-bottom-color: var(--fill-color);\n",
              "    animation:\n",
              "      spin 1s steps(1) infinite;\n",
              "  }\n",
              "\n",
              "  @keyframes spin {\n",
              "    0% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "      border-left-color: var(--fill-color);\n",
              "    }\n",
              "    20% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    30% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    40% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    60% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    80% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "    90% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "  }\n",
              "</style>\n",
              "\n",
              "  <script>\n",
              "    async function quickchart(key) {\n",
              "      const quickchartButtonEl =\n",
              "        document.querySelector('#' + key + ' button');\n",
              "      quickchartButtonEl.disabled = true;  // To prevent multiple clicks.\n",
              "      quickchartButtonEl.classList.add('colab-df-spinner');\n",
              "      try {\n",
              "        const charts = await google.colab.kernel.invokeFunction(\n",
              "            'suggestCharts', [key], {});\n",
              "      } catch (error) {\n",
              "        console.error('Error during call to suggestCharts:', error);\n",
              "      }\n",
              "      quickchartButtonEl.classList.remove('colab-df-spinner');\n",
              "      quickchartButtonEl.classList.add('colab-df-quickchart-complete');\n",
              "    }\n",
              "    (() => {\n",
              "      let quickchartButtonEl =\n",
              "        document.querySelector('#df-be61f96e-b1fa-40f7-af51-cc6005b4c37b button');\n",
              "      quickchartButtonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "    })();\n",
              "  </script>\n",
              "</div>\n",
              "\n",
              "    </div>\n",
              "  </div>\n"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "dataframe",
              "variable_name": "vihsd_train",
              "summary": "{\n  \"name\": \"vihsd_train\",\n  \"rows\": 24048,\n  \"fields\": [\n    {\n      \"column\": \"free_text\",\n      \"properties\": {\n        \"dtype\": \"string\",\n        \"num_unique_values\": 22557,\n        \"samples\": [\n          \"@Qu\\u1ed1c B\\u1ea3o L\\u01b0u l\\u00ean ph\\u01b0\\u1eddng l\\u00e0m vi\\u1ec7c l\\u00e0 bi\\u1ebft li\\u1ec1n nha, \\u0111\\u00f4i co \\u1edf \\u0111\\u00e2y chi cho m\\u1ec7t =]]\",\n          \"C\\u1ea3m th\\u1ea5y ch\\u1eedi l\\u0111ao l\\u00e0 1 th\\u00fa vui x\\u1ea3 stress do b\\u1ecdn n\\u00f3 qu\\u00e1 ngu v\\u00e0 nguy hi\\u1ec3m :)))\",\n          \"C\\u1eadu M\\u1ea1nh cao nh\\u00e2n\"\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    },\n    {\n      \"column\": \"label_id\",\n      \"properties\": {\n        \"dtype\": \"number\",\n        \"std\": 0,\n        \"min\": 0,\n        \"max\": 2,\n        \"num_unique_values\": 3,\n        \"samples\": [\n          0,\n          2,\n          1\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    }\n  ]\n}"
            }
          },
          "metadata": {},
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "combined_train = pd.concat([vihsd_train.rename(columns={'free_text': 'text'}), train.rename(columns={'label': 'label_id'})])\n",
        "combined_dev = pd.concat([vihsd_dev.rename(columns={'free_text': 'text'}), val.rename(columns={'label': 'label_id'})])\n",
        "combined_test = pd.concat([vihsd_test.rename(columns={'free_text': 'text'}), test.rename(columns={'label': 'label_id'})])\n",
        "\n",
        "# Shuffle each combined dataset\n",
        "combined_train = combined_train.sample(frac=1, random_state=42).reset_index(drop=True)\n",
        "combined_dev = combined_dev.sample(frac=1, random_state=42).reset_index(drop=True)\n",
        "combined_test = combined_test.sample(frac=1, random_state=42).reset_index(drop=True)"
      ],
      "metadata": {
        "id": "OmevyfSDophB"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Tập train của 2k7 trước và sau khi bổ sung thêm vihsd"
      ],
      "metadata": {
        "id": "zdfuQWaXpa3h"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "len(train), len(vihsd_train), len(combined_train)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pLn1v_tbpLli",
        "outputId": "d0b028ba-3798-418d-9f9b-51fe306fc663"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(1890, 24048, 25938)"
            ]
          },
          "metadata": {},
          "execution_count": 11
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "combined_train.head()"
      ],
      "metadata": {
        "id": "chX0bletpy0p"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Xóa các row bị NaN"
      ],
      "metadata": {
        "id": "xbsXCXIuYGMX"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "combined_train.info()"
      ],
      "metadata": {
        "id": "Dv0q8fx5W7Go"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "combined_train.dropna(inplace=True)"
      ],
      "metadata": {
        "id": "Qau6iF7-XJv7"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## The distributions of three labels on the train, val, and test sets"
      ],
      "metadata": {
        "id": "vVUi0BLRDWNz"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "label_id_to_name = {0: 'NORMAL', 1: 'HATE', 2: 'HARASSMENT'}\n",
        "\n",
        "train_label_counts = combined_train['label_id'].map(label_id_to_name).value_counts()\n",
        "val_label_counts = combined_dev['label_id'].map(label_id_to_name).value_counts()\n",
        "test_label_counts = combined_test['label_id'].map(label_id_to_name).value_counts()\n",
        "\n",
        "fig, ax = plt.subplots(figsize=(6, 4))\n",
        "\n",
        "# Heights of the bars\n",
        "hate_speech_counts = [train_label_counts.get('HATE', 0), val_label_counts.get('HATE', 0), test_label_counts.get('HATE', 0)]\n",
        "harassment_counts = [train_label_counts.get('HARASSMENT', 0), val_label_counts.get('HARASSMENT', 0), test_label_counts.get('HARASSMENT', 0)]\n",
        "normal_counts = [train_label_counts.get('NORMAL', 0), val_label_counts.get('NORMAL', 0), test_label_counts.get('NORMAL', 0)]\n",
        "\n",
        "# The x position of bars\n",
        "barWidth = 0.25\n",
        "r1 = np.arange(len(hate_speech_counts))\n",
        "r2 = [x + barWidth for x in r1]\n",
        "r3 = [x + barWidth for x in r2]\n",
        "\n",
        "# Plotting the bars\n",
        "plt.bar(r1, hate_speech_counts, color='grey', width=barWidth, edgecolor='grey', label='HATE SPEECH')\n",
        "plt.bar(r2, harassment_counts, color='black', width=barWidth, edgecolor='black', label='HARASSMENT')\n",
        "plt.bar(r3, normal_counts, color='white', width=barWidth, edgecolor='black', hatch='/', label='NORMAL')\n",
        "\n",
        "# Adding the annotations for counts on top of the bars\n",
        "def add_labels(x, counts):\n",
        "    for i in range(len(x)):\n",
        "        plt.text(x[i], counts[i] + max(counts) * 0.01, f'{counts[i]:,.0f}', ha='center')\n",
        "add_labels(r1, hate_speech_counts)\n",
        "add_labels(r2, harassment_counts)\n",
        "add_labels(r3, normal_counts)\n",
        "\n",
        "# General layout\n",
        "plt.xticks([r + barWidth for r in range(len(hate_speech_counts))], ['TRAIN', 'VAL', 'TEST'])\n",
        "plt.ylabel('Number of comments')\n",
        "plt.yscale('log')\n",
        "plt.title('Distribution of Comments by Label (Actual Counts)')\n",
        "plt.legend()\n",
        "\n",
        "plt.grid(which='both', axis='y', linestyle='--', linewidth=0.7)\n",
        "\n",
        "plt.tight_layout()\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 407
        },
        "id": "tajxUrg0Dnrn",
        "outputId": "4f34d57f-e2df-4bbf-a682-2fd45a72ac4b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 600x400 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Biểu đồ cho thấy sự mất cân bằng đáng kể trong tập dữ liệu, với số lượng nhận xét 'normal' vượt xa số lượng nhận xét về 'hate_speech' và 'harassment'.\n",
        "\n",
        "1. **Mất cân bằng**: Các mô hình có xu hướng thiên về class đa số, có khả năng dẫn đến việc xác định kém các class thiểu số.\n",
        "2. **Overfitting**: Dữ liệu hạn chế về 'hate_speech' và 'harassment' có thể khiến mô hình overfitting dữ liệu đào tạo, cản trở khả năng khái quát hóa của mô hình.\n",
        "3. Có thể không có đủ sự đa dạng trong các ví dụ về hate_speech và harassment để nắm bắt được các sắc thái của nội dung đó.\n",
        "4. **Metric**: Accuracy không phải là số liệu đáng tin cậy do tính mất cân bằng nên sẽ dùng điểm F1 mang lại nhiều thông tin hơn."
      ],
      "metadata": {
        "id": "oQprowX8sLvc"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Vietnamese stopwords list\n",
        "STOPWORDS = 'vietnamese-stopwords-dash.txt'\n",
        "with open(DATA_PATH + STOPWORDS, \"r\") as ins:\n",
        "    stopwords = []\n",
        "    for line in ins:\n",
        "        dd = line.strip('\\n')\n",
        "        stopwords.append(dd)\n",
        "    stopwords = set(stopwords)"
      ],
      "metadata": {
        "id": "NZ8phgLx5Je0"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "# Function to generate word clouds\n",
        "def generate_word_cloud(text_series, title):\n",
        "    text = ' '.join(text_series)\n",
        "    wordcloud = WordCloud(stopwords=stopwords, background_color=\"white\", width=800, height=400).generate(text)\n",
        "    plt.figure(figsize=(6, 4))\n",
        "    plt.imshow(wordcloud, interpolation='bilinear')\n",
        "    plt.axis(\"off\")\n",
        "    plt.title(title)\n",
        "    plt.show()\n",
        "\n",
        "# Generate word clouds for each class\n",
        "for label in combined_train['label_id'].unique():\n",
        "    generate_word_cloud(combined_train[combined_train['label_id'] == label]['text'].astype(str), f'Word Cloud for {label_id_to_name[label]} Comments')\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 193
        },
        "id": "J0hB3l_4vxw7",
        "outputId": "5972811a-5604-4397-816e-d1a38f3e1198"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "error",
          "ename": "NameError",
          "evalue": "name 'label_id_to_name' is not defined",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-46-0dbc9be7cf68>\u001b[0m in \u001b[0;36m<cell line: 21>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     20\u001b[0m \u001b[0;31m# Generate word clouds for each class\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     21\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mlabel\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mcombined_train\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'label_id'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0munique\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 22\u001b[0;31m     \u001b[0mgenerate_word_cloud\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcombined_train\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mcombined_train\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'label_id'\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0mlabel\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'text'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mastype\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mstr\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34mf'Word Cloud for {label_id_to_name[label]} Comments'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
            "\u001b[0;31mNameError\u001b[0m: name 'label_id_to_name' is not defined"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "combined_train[(combined_train['label_id'] == 1) & (combined_train['text'].str.contains(' m '))].head()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 206
        },
        "id": "D8-Rpktm17PW",
        "outputId": "0f88722f-f1ee-4238-9152-5e1cdc51c08e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                                   text  label_id\n",
              "598   Thanh Tuyền câm mẹ mồm đi,tốt đb,t bị chúng m ...         1\n",
              "693   Trời má lại đẹp Lan Tường m biết anh này không...         1\n",
              "750   Dăm ba con chó tín dụng đen tụi m dám tới nhà ...         1\n",
              "1276                                     Bố m đéo sợ ok         1\n",
              "1648  :))))) nay thi doc quyen :V d' cho tu nhan tha...         1"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-1502b394-9153-4706-a70b-ccea7db526c1\" class=\"colab-df-container\">\n",
              "    <div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>text</th>\n",
              "      <th>label_id</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>598</th>\n",
              "      <td>Thanh Tuyền câm mẹ mồm đi,tốt đb,t bị chúng m ...</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>693</th>\n",
              "      <td>Trời má lại đẹp Lan Tường m biết anh này không...</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>750</th>\n",
              "      <td>Dăm ba con chó tín dụng đen tụi m dám tới nhà ...</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1276</th>\n",
              "      <td>Bố m đéo sợ ok</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1648</th>\n",
              "      <td>:))))) nay thi doc quyen :V d' cho tu nhan tha...</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "    <div class=\"colab-df-buttons\">\n",
              "\n",
              "  <div class=\"colab-df-container\">\n",
              "    <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-1502b394-9153-4706-a70b-ccea7db526c1')\"\n",
              "            title=\"Convert this dataframe to an interactive table.\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\" viewBox=\"0 -960 960 960\">\n",
              "    <path d=\"M120-120v-720h720v720H120Zm60-500h600v-160H180v160Zm220 220h160v-160H400v160Zm0 220h160v-160H400v160ZM180-400h160v-160H180v160Zm440 0h160v-160H620v160ZM180-180h160v-160H180v160Zm440 0h160v-160H620v160Z\"/>\n",
              "  </svg>\n",
              "    </button>\n",
              "\n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    .colab-df-buttons div {\n",
              "      margin-bottom: 4px;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "    <script>\n",
              "      const buttonEl =\n",
              "        document.querySelector('#df-1502b394-9153-4706-a70b-ccea7db526c1 button.colab-df-convert');\n",
              "      buttonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "      async function convertToInteractive(key) {\n",
              "        const element = document.querySelector('#df-1502b394-9153-4706-a70b-ccea7db526c1');\n",
              "        const dataTable =\n",
              "          await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                    [key], {});\n",
              "        if (!dataTable) return;\n",
              "\n",
              "        const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "          '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "          + ' to learn more about interactive tables.';\n",
              "        element.innerHTML = '';\n",
              "        dataTable['output_type'] = 'display_data';\n",
              "        await google.colab.output.renderOutput(dataTable, element);\n",
              "        const docLink = document.createElement('div');\n",
              "        docLink.innerHTML = docLinkHtml;\n",
              "        element.appendChild(docLink);\n",
              "      }\n",
              "    </script>\n",
              "  </div>\n",
              "\n",
              "\n",
              "<div id=\"df-56aaf92b-a6fe-4929-9f1c-d5cdd1490649\">\n",
              "  <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-56aaf92b-a6fe-4929-9f1c-d5cdd1490649')\"\n",
              "            title=\"Suggest charts\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "<svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "     width=\"24px\">\n",
              "    <g>\n",
              "        <path d=\"M19 3H5c-1.1 0-2 .9-2 2v14c0 1.1.9 2 2 2h14c1.1 0 2-.9 2-2V5c0-1.1-.9-2-2-2zM9 17H7v-7h2v7zm4 0h-2V7h2v10zm4 0h-2v-4h2v4z\"/>\n",
              "    </g>\n",
              "</svg>\n",
              "  </button>\n",
              "\n",
              "<style>\n",
              "  .colab-df-quickchart {\n",
              "      --bg-color: #E8F0FE;\n",
              "      --fill-color: #1967D2;\n",
              "      --hover-bg-color: #E2EBFA;\n",
              "      --hover-fill-color: #174EA6;\n",
              "      --disabled-fill-color: #AAA;\n",
              "      --disabled-bg-color: #DDD;\n",
              "  }\n",
              "\n",
              "  [theme=dark] .colab-df-quickchart {\n",
              "      --bg-color: #3B4455;\n",
              "      --fill-color: #D2E3FC;\n",
              "      --hover-bg-color: #434B5C;\n",
              "      --hover-fill-color: #FFFFFF;\n",
              "      --disabled-bg-color: #3B4455;\n",
              "      --disabled-fill-color: #666;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart {\n",
              "    background-color: var(--bg-color);\n",
              "    border: none;\n",
              "    border-radius: 50%;\n",
              "    cursor: pointer;\n",
              "    display: none;\n",
              "    fill: var(--fill-color);\n",
              "    height: 32px;\n",
              "    padding: 0;\n",
              "    width: 32px;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart:hover {\n",
              "    background-color: var(--hover-bg-color);\n",
              "    box-shadow: 0 1px 2px rgba(60, 64, 67, 0.3), 0 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "    fill: var(--button-hover-fill-color);\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart-complete:disabled,\n",
              "  .colab-df-quickchart-complete:disabled:hover {\n",
              "    background-color: var(--disabled-bg-color);\n",
              "    fill: var(--disabled-fill-color);\n",
              "    box-shadow: none;\n",
              "  }\n",
              "\n",
              "  .colab-df-spinner {\n",
              "    border: 2px solid var(--fill-color);\n",
              "    border-color: transparent;\n",
              "    border-bottom-color: var(--fill-color);\n",
              "    animation:\n",
              "      spin 1s steps(1) infinite;\n",
              "  }\n",
              "\n",
              "  @keyframes spin {\n",
              "    0% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "      border-left-color: var(--fill-color);\n",
              "    }\n",
              "    20% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    30% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    40% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    60% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    80% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "    90% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "  }\n",
              "</style>\n",
              "\n",
              "  <script>\n",
              "    async function quickchart(key) {\n",
              "      const quickchartButtonEl =\n",
              "        document.querySelector('#' + key + ' button');\n",
              "      quickchartButtonEl.disabled = true;  // To prevent multiple clicks.\n",
              "      quickchartButtonEl.classList.add('colab-df-spinner');\n",
              "      try {\n",
              "        const charts = await google.colab.kernel.invokeFunction(\n",
              "            'suggestCharts', [key], {});\n",
              "      } catch (error) {\n",
              "        console.error('Error during call to suggestCharts:', error);\n",
              "      }\n",
              "      quickchartButtonEl.classList.remove('colab-df-spinner');\n",
              "      quickchartButtonEl.classList.add('colab-df-quickchart-complete');\n",
              "    }\n",
              "    (() => {\n",
              "      let quickchartButtonEl =\n",
              "        document.querySelector('#df-56aaf92b-a6fe-4929-9f1c-d5cdd1490649 button');\n",
              "      quickchartButtonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "    })();\n",
              "  </script>\n",
              "</div>\n",
              "\n",
              "    </div>\n",
              "  </div>\n"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "dataframe",
              "summary": "{\n  \"name\": \"combined_train[(combined_train['label_id'] == 1) & (combined_train['text']\",\n  \"rows\": 5,\n  \"fields\": [\n    {\n      \"column\": \"text\",\n      \"properties\": {\n        \"dtype\": \"string\",\n        \"num_unique_values\": 5,\n        \"samples\": [\n          \"Tr\\u1eddi m\\u00e1 l\\u1ea1i \\u0111\\u1eb9p Lan T\\u01b0\\u1eddng m bi\\u1ebft anh n\\u00e0y kh\\u00f4ng ???? M\\u1edf mi\\u1ec7ng kh\\u00f4ng bi\\u1ebft l\\u00e0 v\\u1ea3 cho r\\u1edbt mu lu\\u00f4n nha\",\n          \":))))) nay thi doc quyen :V d' cho tu nhan tham gia om chet me chung m di :v de dan ganh d gi\",\n          \"D\\u0103m ba con ch\\u00f3 t\\u00edn d\\u1ee5ng \\u0111en t\\u1ee5i m d\\u00e1m t\\u1edbi nh\\u00e0 ai? L\\u00e1o lol quen. Ngu th\\u00ec ch\\u1ebft. To\\u00e0n b\\u1ecdn t\\u00e0u kh\\u1ef1a \\u0111\\u00f2i \\u0111e d\\u1ecda d\\u00e2n VN \\u00e0? Tu\\u1ed5i cc\"\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    },\n    {\n      \"column\": \"label_id\",\n      \"properties\": {\n        \"dtype\": \"number\",\n        \"std\": 0,\n        \"min\": 1,\n        \"max\": 1,\n        \"num_unique_values\": 1,\n        \"samples\": [\n          1\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    }\n  ]\n}"
            }
          },
          "metadata": {},
          "execution_count": 18
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "vihsd_train[(vihsd_train['label_id'] == 1) & (vihsd_train['free_text'].str.contains(' m '))].head(20)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 551
        },
        "id": "S90Cc_4n81oG",
        "outputId": "d4251f0a-254f-46b4-e953-431e5a7e73c7"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                               free_text  label_id\n",
              "2115   Nó thích mình nó đã yêu cmm luôn chứ bắt m đợi...         1\n",
              "4406                         PhẠm mạng m lag chứ chậm cc         1\n",
              "8601                          Đông Ca là địt cụ m ngủ đi         1\n",
              "13409     cc, t bắn 3 phát rồi đấy, ngon vcl ? m đui à ?         1\n",
              "14061                                     Bố m đéo sợ ok         1\n",
              "14087                   Vitamin Nhung cđ m chứ ai nữa 😁😁         1\n",
              "14931       Cũg chưa biết tin ai đc m ơi =)) đồn như lời         1\n",
              "15122  Cũng đồng ý vs ad nhưng bạn m từ khi tham gia ...         1\n",
              "15353                  Tụt mẹ m ra cho nhanh đỡ phải mặc         1\n",
              "15667                   Đinhh Thị Hoàii Thuu ngủ mẹ m đi         1\n",
              "17147  đéo phải cái gì thích thử là thử đc nhá,nhất l...         1\n",
              "17263  :))))) nay thi doc quyen :V d' cho tu nhan tha...         1\n",
              "17709  Dm tao còn nhớ lần t đi thi học sinh giỏi lớp ...         1\n",
              "18019                      Ghi thêm chữ đ*t chị m nx :))         1\n",
              "18300  ờ ờ . nợ thì chúng m nợ nhé. :T t đéo có nợ ai...         1\n",
              "20055  Ngọc Hải ê, biểu tượng cl của m thành trend bê...         1"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-6e7c51d4-9b10-4079-aec0-7069502589fe\" class=\"colab-df-container\">\n",
              "    <div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>free_text</th>\n",
              "      <th>label_id</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>2115</th>\n",
              "      <td>Nó thích mình nó đã yêu cmm luôn chứ bắt m đợi...</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4406</th>\n",
              "      <td>PhẠm mạng m lag chứ chậm cc</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>8601</th>\n",
              "      <td>Đông Ca là địt cụ m ngủ đi</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>13409</th>\n",
              "      <td>cc, t bắn 3 phát rồi đấy, ngon vcl ? m đui à ?</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>14061</th>\n",
              "      <td>Bố m đéo sợ ok</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>14087</th>\n",
              "      <td>Vitamin Nhung cđ m chứ ai nữa 😁😁</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>14931</th>\n",
              "      <td>Cũg chưa biết tin ai đc m ơi =)) đồn như lời</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>15122</th>\n",
              "      <td>Cũng đồng ý vs ad nhưng bạn m từ khi tham gia ...</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>15353</th>\n",
              "      <td>Tụt mẹ m ra cho nhanh đỡ phải mặc</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>15667</th>\n",
              "      <td>Đinhh Thị Hoàii Thuu ngủ mẹ m đi</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>17147</th>\n",
              "      <td>đéo phải cái gì thích thử là thử đc nhá,nhất l...</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>17263</th>\n",
              "      <td>:))))) nay thi doc quyen :V d' cho tu nhan tha...</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>17709</th>\n",
              "      <td>Dm tao còn nhớ lần t đi thi học sinh giỏi lớp ...</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>18019</th>\n",
              "      <td>Ghi thêm chữ đ*t chị m nx :))</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>18300</th>\n",
              "      <td>ờ ờ . nợ thì chúng m nợ nhé. :T t đéo có nợ ai...</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>20055</th>\n",
              "      <td>Ngọc Hải ê, biểu tượng cl của m thành trend bê...</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "    <div class=\"colab-df-buttons\">\n",
              "\n",
              "  <div class=\"colab-df-container\">\n",
              "    <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-6e7c51d4-9b10-4079-aec0-7069502589fe')\"\n",
              "            title=\"Convert this dataframe to an interactive table.\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\" viewBox=\"0 -960 960 960\">\n",
              "    <path d=\"M120-120v-720h720v720H120Zm60-500h600v-160H180v160Zm220 220h160v-160H400v160Zm0 220h160v-160H400v160ZM180-400h160v-160H180v160Zm440 0h160v-160H620v160ZM180-180h160v-160H180v160Zm440 0h160v-160H620v160Z\"/>\n",
              "  </svg>\n",
              "    </button>\n",
              "\n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    .colab-df-buttons div {\n",
              "      margin-bottom: 4px;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "    <script>\n",
              "      const buttonEl =\n",
              "        document.querySelector('#df-6e7c51d4-9b10-4079-aec0-7069502589fe button.colab-df-convert');\n",
              "      buttonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "      async function convertToInteractive(key) {\n",
              "        const element = document.querySelector('#df-6e7c51d4-9b10-4079-aec0-7069502589fe');\n",
              "        const dataTable =\n",
              "          await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                    [key], {});\n",
              "        if (!dataTable) return;\n",
              "\n",
              "        const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "          '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "          + ' to learn more about interactive tables.';\n",
              "        element.innerHTML = '';\n",
              "        dataTable['output_type'] = 'display_data';\n",
              "        await google.colab.output.renderOutput(dataTable, element);\n",
              "        const docLink = document.createElement('div');\n",
              "        docLink.innerHTML = docLinkHtml;\n",
              "        element.appendChild(docLink);\n",
              "      }\n",
              "    </script>\n",
              "  </div>\n",
              "\n",
              "\n",
              "<div id=\"df-f5a840c6-c096-47b3-8571-5f10cba95e0b\">\n",
              "  <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-f5a840c6-c096-47b3-8571-5f10cba95e0b')\"\n",
              "            title=\"Suggest charts\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "<svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "     width=\"24px\">\n",
              "    <g>\n",
              "        <path d=\"M19 3H5c-1.1 0-2 .9-2 2v14c0 1.1.9 2 2 2h14c1.1 0 2-.9 2-2V5c0-1.1-.9-2-2-2zM9 17H7v-7h2v7zm4 0h-2V7h2v10zm4 0h-2v-4h2v4z\"/>\n",
              "    </g>\n",
              "</svg>\n",
              "  </button>\n",
              "\n",
              "<style>\n",
              "  .colab-df-quickchart {\n",
              "      --bg-color: #E8F0FE;\n",
              "      --fill-color: #1967D2;\n",
              "      --hover-bg-color: #E2EBFA;\n",
              "      --hover-fill-color: #174EA6;\n",
              "      --disabled-fill-color: #AAA;\n",
              "      --disabled-bg-color: #DDD;\n",
              "  }\n",
              "\n",
              "  [theme=dark] .colab-df-quickchart {\n",
              "      --bg-color: #3B4455;\n",
              "      --fill-color: #D2E3FC;\n",
              "      --hover-bg-color: #434B5C;\n",
              "      --hover-fill-color: #FFFFFF;\n",
              "      --disabled-bg-color: #3B4455;\n",
              "      --disabled-fill-color: #666;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart {\n",
              "    background-color: var(--bg-color);\n",
              "    border: none;\n",
              "    border-radius: 50%;\n",
              "    cursor: pointer;\n",
              "    display: none;\n",
              "    fill: var(--fill-color);\n",
              "    height: 32px;\n",
              "    padding: 0;\n",
              "    width: 32px;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart:hover {\n",
              "    background-color: var(--hover-bg-color);\n",
              "    box-shadow: 0 1px 2px rgba(60, 64, 67, 0.3), 0 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "    fill: var(--button-hover-fill-color);\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart-complete:disabled,\n",
              "  .colab-df-quickchart-complete:disabled:hover {\n",
              "    background-color: var(--disabled-bg-color);\n",
              "    fill: var(--disabled-fill-color);\n",
              "    box-shadow: none;\n",
              "  }\n",
              "\n",
              "  .colab-df-spinner {\n",
              "    border: 2px solid var(--fill-color);\n",
              "    border-color: transparent;\n",
              "    border-bottom-color: var(--fill-color);\n",
              "    animation:\n",
              "      spin 1s steps(1) infinite;\n",
              "  }\n",
              "\n",
              "  @keyframes spin {\n",
              "    0% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "      border-left-color: var(--fill-color);\n",
              "    }\n",
              "    20% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    30% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    40% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    60% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    80% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "    90% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "  }\n",
              "</style>\n",
              "\n",
              "  <script>\n",
              "    async function quickchart(key) {\n",
              "      const quickchartButtonEl =\n",
              "        document.querySelector('#' + key + ' button');\n",
              "      quickchartButtonEl.disabled = true;  // To prevent multiple clicks.\n",
              "      quickchartButtonEl.classList.add('colab-df-spinner');\n",
              "      try {\n",
              "        const charts = await google.colab.kernel.invokeFunction(\n",
              "            'suggestCharts', [key], {});\n",
              "      } catch (error) {\n",
              "        console.error('Error during call to suggestCharts:', error);\n",
              "      }\n",
              "      quickchartButtonEl.classList.remove('colab-df-spinner');\n",
              "      quickchartButtonEl.classList.add('colab-df-quickchart-complete');\n",
              "    }\n",
              "    (() => {\n",
              "      let quickchartButtonEl =\n",
              "        document.querySelector('#df-f5a840c6-c096-47b3-8571-5f10cba95e0b button');\n",
              "      quickchartButtonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "    })();\n",
              "  </script>\n",
              "</div>\n",
              "\n",
              "    </div>\n",
              "  </div>\n"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "dataframe",
              "summary": "{\n  \"name\": \"vihsd_train[(vihsd_train['label_id'] == 1) & (vihsd_train['free_text']\",\n  \"rows\": 16,\n  \"fields\": [\n    {\n      \"column\": \"free_text\",\n      \"properties\": {\n        \"dtype\": \"string\",\n        \"num_unique_values\": 16,\n        \"samples\": [\n          \"N\\u00f3 th\\u00edch m\\u00ecnh n\\u00f3 \\u0111\\u00e3 y\\u00eau cmm lu\\u00f4n ch\\u1ee9 b\\u1eaft m \\u0111\\u1ee3i l\\u00e0m l\\u1ed3n g\\u00ec. \\u0110\\u1ee3i ch\\u1ec9 l\\u00e0 l\\u00ed do \\u0111\\u1ec3 n\\u00f3 t\\u00ecm xem c\\u00f3 k\\u00e8o n\\u00e0o ngon h\\u01a1n kh\\u00f4ng th\\u00f4i\",\n          \"Ph\\u1ea0m m\\u1ea1ng m lag ch\\u1ee9 ch\\u1eadm cc\",\n          \"Vitamin Nhung c\\u0111 m ch\\u1ee9 ai n\\u1eefa \\ud83d\\ude01\\ud83d\\ude01\"\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    },\n    {\n      \"column\": \"label_id\",\n      \"properties\": {\n        \"dtype\": \"number\",\n        \"std\": 0,\n        \"min\": 1,\n        \"max\": 1,\n        \"num_unique_values\": 1,\n        \"samples\": [\n          1\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    }\n  ]\n}"
            }
          },
          "metadata": {},
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Modeling"
      ],
      "metadata": {
        "id": "7RKBwWy697AI"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Preprocessing\n",
        "\n"
      ],
      "metadata": {
        "id": "dOy1SIvACU98"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "\n",
        "\n",
        "1.   Word-segmentating texts into words by the pyvi tool\n",
        "2.    Removing stopwords\n",
        "3. Changing all texts into lower cases\n",
        "4. Removing special characters such as hashtags, urls, and mention tags.\n",
        "\n",
        "\n",
        "\n",
        "\n"
      ],
      "metadata": {
        "id": "oBfVehbhEI1J"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "def filter_stop_words(train_sentences, stop_words):\n",
        "    new_sent = [word for word in train_sentences.split() if word not in stop_words]\n",
        "    train_sentences = ' '.join(new_sent)\n",
        "\n",
        "    return train_sentences\n",
        "\n",
        "def remove_special_characters(text):\n",
        "    # Remove URLs\n",
        "    text = re.sub(r'http[s]?://(?:[a-zA-Z]|[0-9]|[$-_@.&+]|[!*\\\\(\\\\),]|(?:%[0-9a-fA-F][0-9a-fA-F]))+', '', text)\n",
        "    # Remove mentions\n",
        "    text = re.sub(r'@\\w+', '', text)\n",
        "    # Remove hashtags\n",
        "    text = re.sub(r'#\\w+', '', text)\n",
        "    return text\n",
        "\n",
        "def preprocess(text, tokenized = True, lowercased = True):\n",
        "    text = ViTokenizer.tokenize(text) if tokenized else text\n",
        "    text = filter_stop_words(text, stopwords)\n",
        "    text = remove_special_characters(text)\n",
        "    text = emoji.replace_emoji(text,\"\")\n",
        "    text = text.lower() if lowercased else text\n",
        "    return text\n",
        "\n",
        "def pre_process_features(X, y, tokenized=True, lowercased=True):\n",
        "    X_processed = [preprocess(str(p), tokenized=tokenized, lowercased=lowercased) for p in list(X)]\n",
        "    non_empty_indices = [idx for idx, ele in enumerate(X_processed) if ele]\n",
        "    X_processed = pd.Series(X_processed)[non_empty_indices].tolist()\n",
        "    y_processed = y.iloc[non_empty_indices].tolist()\n",
        "    return X_processed, y_processed\n"
      ],
      "metadata": {
        "id": "HCGZjflDCbRy"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "train_X, train_y = pre_process_features(combined_train['text'], combined_train['label_id'], tokenized=True, lowercased = True)\n",
        "dev_X, dev_y = pre_process_features(combined_dev['text'], combined_dev['label_id'], tokenized=True, lowercased = True)\n",
        "test_X, test_y = pre_process_features(combined_test['text'], combined_test['label_id'], tokenized=True, lowercased = True)"
      ],
      "metadata": {
        "id": "KkbvDajmfRpb"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Tập train trước và sau khi pre_process"
      ],
      "metadata": {
        "id": "VEXZ-w9EvgV0"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "train_X[:10]"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "h3P212upgwGh",
        "outputId": "6fa6cd41-540e-4cc7-cf3f-1b1535e03dc5"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['Mn review gì đó như kiểu phim hay qtqd , các kiểu đi để mình còn hi_vọng 🤔',\n",
              " 'Cho mình xin đc nhé',\n",
              " 'Sống chậm thôi các má = ) )',\n",
              " 'Ai cho em cái link của bộ_tộc mixi gaming với em tìm không thấy : ( (',\n",
              " 'Haizzz_Nhung thì đáng trách thật nhưng tình_hình như này thì Việt_Nam vẫn có_thể tiếp_nhận thêm bệnh_nhân từ nước_ngoài về ( (',\n",
              " 'Ma ba rat la xinh xan con noi thiet',\n",
              " 'Hài ...!!',\n",
              " 'Thằng ad \" phản_động \" chỉ giỏi \" xuyên_tạc đường nối chính_sách của đãng và nhà lước , bôi_nhọ các anh_hùng nực nượng vũ_trang nhân_dân \" .Hết đặt mìn xe_đò thì đặt mìn bưu_điện , ném lựu_đạn vào chợ vào rạp_hát , pháo_kích vào nhà dân vào trường_học ko phải là khủng_bố khủng mẹ gì hết , phải gọi là \" đi nàm kách mệnh \" mới đúng .',\n",
              " 'Hóng quá đa : v',\n",
              " 'Phục_vụ giải_trí cho cả đại_đội múa cột giỏi chịu đòn tốt chuyên mở mini show mừng đảng mừng xuân mừng đất_nước đổi_mới : ) )']"
            ]
          },
          "metadata": {},
          "execution_count": 45
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "pd.set_option('max_colwidth',800)"
      ],
      "metadata": {
        "id": "Od9p91cRv28y"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "combined_train.head(10)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 449
        },
        "id": "RpqpTsusnhDl",
        "outputId": "687fd569-dc0c-4dd0-d9be-4b85482d1eea"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                                                                                                                                                                                                                                                                                                                 text  \\\n",
              "0                                                                                                                                                                                                                                                           Mn review gì đó như kiểu phim hay qtqd, các kiểu đi để mình còn hi vọng 🤔   \n",
              "1                                                                                                                                                                                                                                                                                                               Cho mình xin đc nhé   \n",
              "2                                                                                                                                                                                                                                                                                                           Sống chậm thôi các má =))   \n",
              "3                                                                                                                                                                                                                                                                 Ai cho em cái link của bộ tộc mixi gaming với em tìm không thấy :((   \n",
              "4                                                                                                                                                                                                       Haizzz Nhung thì đáng trách thật nhưng tình hình như này thì Việt Nam vẫn có thể tiếp nhận thêm bệnh nhân từ nước ngoài về ((   \n",
              "5                                                                                                                                                                                                                                                                                                 Ma ba rat la xinh xan con noi thiet   \n",
              "6                                                                                                                                                                                                                                                                                                                           Hài ...!!   \n",
              "7  Thằng ad \"phản động\" chỉ giỏi \"xuyên tạc đường nối chính sách của đãng và nhà lước, bôi nhọ các anh hùng nực nượng vũ trang nhân dân\". Hết đặt mìn xe đò thì đặt mìn bưu điện, ném lựu đạn vào chợ vào rạp hát, pháo kích vào nhà dân vào trường học ko phải là khủng bố khủng mẹ gì hết, phải gọi là \"đi nàm kách mệnh\" mới đúng.   \n",
              "8                                                                                                                                                                                                                                                                                                                      Hóng quá đa :v   \n",
              "9                                                                                                                                                                                                          Phục vụ giải trí cho cả đại đội múa cột giỏi chịu đòn tốt chuyên mở mini show mừng đảng mừng xuân mừng đất nước đổi mới:))   \n",
              "\n",
              "   label_id  \n",
              "0         0  \n",
              "1         0  \n",
              "2         0  \n",
              "3         0  \n",
              "4         0  \n",
              "5         0  \n",
              "6         0  \n",
              "7         2  \n",
              "8         0  \n",
              "9         0  "
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-c407e6a5-c36b-4472-b3bb-df7dece1e5ec\" class=\"colab-df-container\">\n",
              "    <div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>text</th>\n",
              "      <th>label_id</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Mn review gì đó như kiểu phim hay qtqd, các kiểu đi để mình còn hi vọng 🤔</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>Cho mình xin đc nhé</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>Sống chậm thôi các má =))</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>Ai cho em cái link của bộ tộc mixi gaming với em tìm không thấy :((</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>Haizzz Nhung thì đáng trách thật nhưng tình hình như này thì Việt Nam vẫn có thể tiếp nhận thêm bệnh nhân từ nước ngoài về ((</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5</th>\n",
              "      <td>Ma ba rat la xinh xan con noi thiet</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>6</th>\n",
              "      <td>Hài ...!!</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>7</th>\n",
              "      <td>Thằng ad \"phản động\" chỉ giỏi \"xuyên tạc đường nối chính sách của đãng và nhà lước, bôi nhọ các anh hùng nực nượng vũ trang nhân dân\". Hết đặt mìn xe đò thì đặt mìn bưu điện, ném lựu đạn vào chợ vào rạp hát, pháo kích vào nhà dân vào trường học ko phải là khủng bố khủng mẹ gì hết, phải gọi là \"đi nàm kách mệnh\" mới đúng.</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>8</th>\n",
              "      <td>Hóng quá đa :v</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9</th>\n",
              "      <td>Phục vụ giải trí cho cả đại đội múa cột giỏi chịu đòn tốt chuyên mở mini show mừng đảng mừng xuân mừng đất nước đổi mới:))</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "    <div class=\"colab-df-buttons\">\n",
              "\n",
              "  <div class=\"colab-df-container\">\n",
              "    <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-c407e6a5-c36b-4472-b3bb-df7dece1e5ec')\"\n",
              "            title=\"Convert this dataframe to an interactive table.\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\" viewBox=\"0 -960 960 960\">\n",
              "    <path d=\"M120-120v-720h720v720H120Zm60-500h600v-160H180v160Zm220 220h160v-160H400v160Zm0 220h160v-160H400v160ZM180-400h160v-160H180v160Zm440 0h160v-160H620v160ZM180-180h160v-160H180v160Zm440 0h160v-160H620v160Z\"/>\n",
              "  </svg>\n",
              "    </button>\n",
              "\n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    .colab-df-buttons div {\n",
              "      margin-bottom: 4px;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "    <script>\n",
              "      const buttonEl =\n",
              "        document.querySelector('#df-c407e6a5-c36b-4472-b3bb-df7dece1e5ec button.colab-df-convert');\n",
              "      buttonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "      async function convertToInteractive(key) {\n",
              "        const element = document.querySelector('#df-c407e6a5-c36b-4472-b3bb-df7dece1e5ec');\n",
              "        const dataTable =\n",
              "          await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                    [key], {});\n",
              "        if (!dataTable) return;\n",
              "\n",
              "        const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "          '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "          + ' to learn more about interactive tables.';\n",
              "        element.innerHTML = '';\n",
              "        dataTable['output_type'] = 'display_data';\n",
              "        await google.colab.output.renderOutput(dataTable, element);\n",
              "        const docLink = document.createElement('div');\n",
              "        docLink.innerHTML = docLinkHtml;\n",
              "        element.appendChild(docLink);\n",
              "      }\n",
              "    </script>\n",
              "  </div>\n",
              "\n",
              "\n",
              "<div id=\"df-83967025-e054-40b8-abdc-f27a03cd69bf\">\n",
              "  <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-83967025-e054-40b8-abdc-f27a03cd69bf')\"\n",
              "            title=\"Suggest charts\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "<svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "     width=\"24px\">\n",
              "    <g>\n",
              "        <path d=\"M19 3H5c-1.1 0-2 .9-2 2v14c0 1.1.9 2 2 2h14c1.1 0 2-.9 2-2V5c0-1.1-.9-2-2-2zM9 17H7v-7h2v7zm4 0h-2V7h2v10zm4 0h-2v-4h2v4z\"/>\n",
              "    </g>\n",
              "</svg>\n",
              "  </button>\n",
              "\n",
              "<style>\n",
              "  .colab-df-quickchart {\n",
              "      --bg-color: #E8F0FE;\n",
              "      --fill-color: #1967D2;\n",
              "      --hover-bg-color: #E2EBFA;\n",
              "      --hover-fill-color: #174EA6;\n",
              "      --disabled-fill-color: #AAA;\n",
              "      --disabled-bg-color: #DDD;\n",
              "  }\n",
              "\n",
              "  [theme=dark] .colab-df-quickchart {\n",
              "      --bg-color: #3B4455;\n",
              "      --fill-color: #D2E3FC;\n",
              "      --hover-bg-color: #434B5C;\n",
              "      --hover-fill-color: #FFFFFF;\n",
              "      --disabled-bg-color: #3B4455;\n",
              "      --disabled-fill-color: #666;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart {\n",
              "    background-color: var(--bg-color);\n",
              "    border: none;\n",
              "    border-radius: 50%;\n",
              "    cursor: pointer;\n",
              "    display: none;\n",
              "    fill: var(--fill-color);\n",
              "    height: 32px;\n",
              "    padding: 0;\n",
              "    width: 32px;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart:hover {\n",
              "    background-color: var(--hover-bg-color);\n",
              "    box-shadow: 0 1px 2px rgba(60, 64, 67, 0.3), 0 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "    fill: var(--button-hover-fill-color);\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart-complete:disabled,\n",
              "  .colab-df-quickchart-complete:disabled:hover {\n",
              "    background-color: var(--disabled-bg-color);\n",
              "    fill: var(--disabled-fill-color);\n",
              "    box-shadow: none;\n",
              "  }\n",
              "\n",
              "  .colab-df-spinner {\n",
              "    border: 2px solid var(--fill-color);\n",
              "    border-color: transparent;\n",
              "    border-bottom-color: var(--fill-color);\n",
              "    animation:\n",
              "      spin 1s steps(1) infinite;\n",
              "  }\n",
              "\n",
              "  @keyframes spin {\n",
              "    0% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "      border-left-color: var(--fill-color);\n",
              "    }\n",
              "    20% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    30% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    40% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    60% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    80% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "    90% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "  }\n",
              "</style>\n",
              "\n",
              "  <script>\n",
              "    async function quickchart(key) {\n",
              "      const quickchartButtonEl =\n",
              "        document.querySelector('#' + key + ' button');\n",
              "      quickchartButtonEl.disabled = true;  // To prevent multiple clicks.\n",
              "      quickchartButtonEl.classList.add('colab-df-spinner');\n",
              "      try {\n",
              "        const charts = await google.colab.kernel.invokeFunction(\n",
              "            'suggestCharts', [key], {});\n",
              "      } catch (error) {\n",
              "        console.error('Error during call to suggestCharts:', error);\n",
              "      }\n",
              "      quickchartButtonEl.classList.remove('colab-df-spinner');\n",
              "      quickchartButtonEl.classList.add('colab-df-quickchart-complete');\n",
              "    }\n",
              "    (() => {\n",
              "      let quickchartButtonEl =\n",
              "        document.querySelector('#df-83967025-e054-40b8-abdc-f27a03cd69bf button');\n",
              "      quickchartButtonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "    })();\n",
              "  </script>\n",
              "</div>\n",
              "\n",
              "    </div>\n",
              "  </div>\n"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "dataframe",
              "variable_name": "combined_train",
              "summary": "{\n  \"name\": \"combined_train\",\n  \"rows\": 25936,\n  \"fields\": [\n    {\n      \"column\": \"text\",\n      \"properties\": {\n        \"dtype\": \"string\",\n        \"num_unique_values\": 24369,\n        \"samples\": [\n          \"Ph\\u00ea qu\\u00e1 ho\\u00e1 \\u0111i\\u00ean\\ud83e\\udd23\",\n          \"\\u0110\\u1eddi v\\u00f4 th\\u01b0\\u1eddng l\\u1eafm..s\\u1ed1ng ph\\u1ea3i tu t\\u00e2m t\\u00edch \\u0111\\u1ee9c c\\u00f4 oi..kh\\u1ea9u nghi\\u1ec7p qu\\u00e1 c\\u00f4 \\u01a1i\",\n          \"Xa\\u0323o l\\u00f4\\u0300n ma\\u0300y bu\\u0301 l\\u1ed3n nhi\\u00ea\\u0300u qua\\u0301 n\\u00ean nga\\u0301o ha\\u0309 th\\u0103\\u0300ng cho\\u0301\"\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    },\n    {\n      \"column\": \"label_id\",\n      \"properties\": {\n        \"dtype\": \"number\",\n        \"std\": 0,\n        \"min\": 0,\n        \"max\": 2,\n        \"num_unique_values\": 3,\n        \"samples\": [\n          0,\n          2,\n          1\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    }\n  ]\n}"
            }
          },
          "metadata": {},
          "execution_count": 24
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Fasttext Embedding"
      ],
      "metadata": {
        "id": "--S1bg3Twm2J"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "EMBEDDING_FILE = 'cc.vi.300.vec'\n",
        "embeddings_index = {}\n",
        "with open(DATA_PATH + EMBEDDING_FILE, encoding='utf8') as f:\n",
        "    for line in f:\n",
        "        values = line.rstrip().rsplit(' ')\n",
        "        word = values[0]\n",
        "        coefs = np.asarray(values[1:], dtype='float32')\n",
        "        embeddings_index[word] = coefs"
      ],
      "metadata": {
        "id": "klr9UpWGwrqk"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "vocabulary_size = 10000\n",
        "embedding_dim = 300\n",
        "\n",
        "tokenizer = text.Tokenizer(lower=False, filters='!\"#$%&()*+,-./:;<=>?@[\\\\]^`{|}~\\t\\n')\n",
        "tokenizer.fit_on_texts(train_X)\n",
        "\n",
        "word_index = tokenizer.word_index\n",
        "# num_words = min(vocabulary_size, len(word_index) + 1)\n",
        "num_words = len(word_index) + 1\n",
        "embedding_matrix = np.zeros((num_words, embedding_dim))\n",
        "\n",
        "for word, i in word_index.items():\n",
        "    if i >= vocabulary_size:\n",
        "        continue\n",
        "\n",
        "    embedding_vector = embeddings_index.get(word)\n",
        "    if embedding_vector is not None:\n",
        "        embedding_matrix[i] = embedding_vector\n"
      ],
      "metadata": {
        "id": "QFNbZ7q5yIxu"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "sequence_length = 100\n",
        "\n",
        "def make_featues(X, y, tokenizer, is_one_hot_label=True):\n",
        "    X = tokenizer.texts_to_sequences(X)\n",
        "    X = sequence.pad_sequences(X, maxlen=sequence_length)\n",
        "    if is_one_hot_label:\n",
        "        y = to_categorical(y, num_classes=3)\n",
        "    return X, y\n",
        "\n",
        "# Make features\n",
        "train_X, train_y = make_featues(train_X, train_y, tokenizer)\n",
        "dev_X, dev_y = make_featues(dev_X, dev_y, tokenizer)\n",
        "test_X, test_y = make_featues(test_X, test_y, tokenizer, is_one_hot_label=False)"
      ],
      "metadata": {
        "id": "XP18q3x929Gd"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "train_X[:2]"
      ],
      "metadata": {
        "id": "P-85hpA65AU-",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1810be8a-3c81-494d-a14a-a7dd09c4012e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[   0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,\n",
              "           0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,\n",
              "           0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,\n",
              "           0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,\n",
              "           0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,\n",
              "           0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,\n",
              "           0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,\n",
              "           0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,\n",
              "           0,    0,    0,    0,  301, 1754,   87,  139, 9857,   87,    1,\n",
              "        1239],\n",
              "       [   0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,\n",
              "           0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,\n",
              "           0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,\n",
              "           0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,\n",
              "           0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,\n",
              "           0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,\n",
              "           0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,\n",
              "           0,    0,    0,    0,    0,    0,    0,    0,    0,    0,    0,\n",
              "           0,    0,    0,    0,    0,    0,    0,    0,    0,    0,  116,\n",
              "           6]], dtype=int32)"
            ]
          },
          "metadata": {},
          "execution_count": 28
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "train_y[:2]"
      ],
      "metadata": {
        "id": "o7Uoog8l5JVR",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "aaee318f-abc0-4ac0-9715-753ff94e8138"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[1., 0., 0.],\n",
              "       [1., 0., 0.]], dtype=float32)"
            ]
          },
          "metadata": {},
          "execution_count": 29
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Text-CNN"
      ],
      "metadata": {
        "id": "dKDhkGRmHWVq"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "batch_size = 256\n",
        "epochs = 40\n",
        "drop = 0.5\n",
        "\n",
        "filter_sizes = [2,3,5]\n",
        "num_filters = 32\n",
        "\n",
        "inputs = Input(shape=(sequence_length,), dtype='int32')\n",
        "embedding = Embedding(input_dim=num_words, output_dim=embedding_dim, input_length=sequence_length, weights=[embedding_matrix])(inputs)\n",
        "reshape = Reshape((sequence_length,embedding_dim,1))(embedding)\n",
        "\n",
        "conv_0 = Conv2D(num_filters, kernel_size=(filter_sizes[0], embedding_dim), padding='valid', kernel_initializer='normal', activation='elu')(reshape)\n",
        "conv_1 = Conv2D(num_filters, kernel_size=(filter_sizes[1], embedding_dim), padding='valid', kernel_initializer='normal', activation='elu')(reshape)\n",
        "conv_2 = Conv2D(num_filters, kernel_size=(filter_sizes[2], embedding_dim), padding='valid', kernel_initializer='normal', activation='elu')(reshape)\n",
        "\n",
        "maxpool_0 = MaxPool2D(pool_size=(sequence_length - filter_sizes[0] + 1, 1), strides=(1,1), padding='valid')(conv_0)\n",
        "maxpool_1 = MaxPool2D(pool_size=(sequence_length - filter_sizes[1] + 1, 1), strides=(1,1), padding='valid')(conv_1)\n",
        "maxpool_2 = MaxPool2D(pool_size=(sequence_length - filter_sizes[2] + 1, 1), strides=(1,1), padding='valid')(conv_2)\n",
        "\n",
        "concatenated_tensor = Concatenate(axis=1)([maxpool_0, maxpool_1, maxpool_2])\n",
        "flatten = Flatten()(concatenated_tensor)\n",
        "dropout = Dropout(drop)(flatten)\n",
        "output = Dense(units=3, activation='softmax')(dropout)\n",
        "\n",
        "# this creates a model that includes\n",
        "model = Model(inputs=inputs, outputs=output)\n",
        "\n",
        "model.summary()\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "6SnB50O4HVhx",
        "outputId": "7fc8ce97-ef48-4239-8df0-b79cac9a290d"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                Output Shape                 Param #   Connected to                  \n",
            "==================================================================================================\n",
            " input_1 (InputLayer)        [(None, 100)]                0         []                            \n",
            "                                                                                                  \n",
            " embedding (Embedding)       (None, 100, 300)             6449700   ['input_1[0][0]']             \n",
            "                                                                                                  \n",
            " reshape (Reshape)           (None, 100, 300, 1)          0         ['embedding[0][0]']           \n",
            "                                                                                                  \n",
            " conv2d (Conv2D)             (None, 99, 1, 32)            19232     ['reshape[0][0]']             \n",
            "                                                                                                  \n",
            " conv2d_1 (Conv2D)           (None, 98, 1, 32)            28832     ['reshape[0][0]']             \n",
            "                                                                                                  \n",
            " conv2d_2 (Conv2D)           (None, 96, 1, 32)            48032     ['reshape[0][0]']             \n",
            "                                                                                                  \n",
            " max_pooling2d (MaxPooling2  (None, 1, 1, 32)             0         ['conv2d[0][0]']              \n",
            " D)                                                                                               \n",
            "                                                                                                  \n",
            " max_pooling2d_1 (MaxPoolin  (None, 1, 1, 32)             0         ['conv2d_1[0][0]']            \n",
            " g2D)                                                                                             \n",
            "                                                                                                  \n",
            " max_pooling2d_2 (MaxPoolin  (None, 1, 1, 32)             0         ['conv2d_2[0][0]']            \n",
            " g2D)                                                                                             \n",
            "                                                                                                  \n",
            " concatenate (Concatenate)   (None, 3, 1, 32)             0         ['max_pooling2d[0][0]',       \n",
            "                                                                     'max_pooling2d_1[0][0]',     \n",
            "                                                                     'max_pooling2d_2[0][0]']     \n",
            "                                                                                                  \n",
            " flatten (Flatten)           (None, 96)                   0         ['concatenate[0][0]']         \n",
            "                                                                                                  \n",
            " dropout (Dropout)           (None, 96)                   0         ['flatten[0][0]']             \n",
            "                                                                                                  \n",
            " dense (Dense)               (None, 3)                    291       ['dropout[0][0]']             \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 6546087 (24.97 MB)\n",
            "Trainable params: 6546087 (24.97 MB)\n",
            "Non-trainable params: 0 (0.00 Byte)\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "MODEL_FILE = 'Text_CNN_model_v1.h5'\n",
        "adam = Adam(learning_rate=1e-4, beta_1=0.9, beta_2=0.999, epsilon=1e-08)\n",
        "\n",
        "model.compile(optimizer=adam, loss='categorical_crossentropy', metrics=['accuracy'])\n",
        "\n",
        "model.fit(train_X, train_y, batch_size=batch_size, epochs=epochs, verbose=1,\n",
        "          validation_data=(dev_X, dev_y))\n",
        "model.save(DATA_PATH + MODEL_FILE)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "BZTiEjmCNNvT",
        "outputId": "4d367d83-75bd-4a9b-c771-ea361292af52"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/40\n",
            "101/101 [==============================] - 40s 277ms/step - loss: 0.8129 - accuracy: 0.7685 - val_loss: 0.6768 - val_accuracy: 0.8151\n",
            "Epoch 2/40\n",
            "101/101 [==============================] - 10s 103ms/step - loss: 0.6158 - accuracy: 0.8236 - val_loss: 0.5659 - val_accuracy: 0.8157\n",
            "Epoch 3/40\n",
            "101/101 [==============================] - 9s 86ms/step - loss: 0.5278 - accuracy: 0.8284 - val_loss: 0.5077 - val_accuracy: 0.8191\n",
            "Epoch 4/40\n",
            "101/101 [==============================] - 6s 53ms/step - loss: 0.4803 - accuracy: 0.8360 - val_loss: 0.4785 - val_accuracy: 0.8294\n",
            "Epoch 5/40\n",
            "101/101 [==============================] - 4s 44ms/step - loss: 0.4468 - accuracy: 0.8441 - val_loss: 0.4568 - val_accuracy: 0.8362\n",
            "Epoch 6/40\n",
            "101/101 [==============================] - 4s 41ms/step - loss: 0.4191 - accuracy: 0.8505 - val_loss: 0.4384 - val_accuracy: 0.8420\n",
            "Epoch 7/40\n",
            "101/101 [==============================] - 4s 36ms/step - loss: 0.3967 - accuracy: 0.8575 - val_loss: 0.4252 - val_accuracy: 0.8447\n",
            "Epoch 8/40\n",
            "101/101 [==============================] - 3s 34ms/step - loss: 0.3735 - accuracy: 0.8642 - val_loss: 0.4143 - val_accuracy: 0.8460\n",
            "Epoch 9/40\n",
            "101/101 [==============================] - 3s 34ms/step - loss: 0.3541 - accuracy: 0.8714 - val_loss: 0.4060 - val_accuracy: 0.8478\n",
            "Epoch 10/40\n",
            "101/101 [==============================] - 4s 35ms/step - loss: 0.3357 - accuracy: 0.8755 - val_loss: 0.3989 - val_accuracy: 0.8488\n",
            "Epoch 11/40\n",
            "101/101 [==============================] - 4s 35ms/step - loss: 0.3168 - accuracy: 0.8830 - val_loss: 0.3938 - val_accuracy: 0.8515\n",
            "Epoch 12/40\n",
            "101/101 [==============================] - 2s 22ms/step - loss: 0.3000 - accuracy: 0.8896 - val_loss: 0.3886 - val_accuracy: 0.8542\n",
            "Epoch 13/40\n",
            "101/101 [==============================] - 3s 28ms/step - loss: 0.2872 - accuracy: 0.8922 - val_loss: 0.3854 - val_accuracy: 0.8559\n",
            "Epoch 14/40\n",
            "101/101 [==============================] - 3s 32ms/step - loss: 0.2709 - accuracy: 0.8982 - val_loss: 0.3823 - val_accuracy: 0.8549\n",
            "Epoch 15/40\n",
            "101/101 [==============================] - 2s 22ms/step - loss: 0.2582 - accuracy: 0.9040 - val_loss: 0.3801 - val_accuracy: 0.8573\n",
            "Epoch 16/40\n",
            "101/101 [==============================] - 3s 28ms/step - loss: 0.2448 - accuracy: 0.9098 - val_loss: 0.3797 - val_accuracy: 0.8587\n",
            "Epoch 17/40\n",
            "101/101 [==============================] - 2s 24ms/step - loss: 0.2310 - accuracy: 0.9135 - val_loss: 0.3788 - val_accuracy: 0.8587\n",
            "Epoch 18/40\n",
            "101/101 [==============================] - 2s 24ms/step - loss: 0.2209 - accuracy: 0.9196 - val_loss: 0.3801 - val_accuracy: 0.8593\n",
            "Epoch 19/40\n",
            "101/101 [==============================] - 2s 25ms/step - loss: 0.2098 - accuracy: 0.9224 - val_loss: 0.3792 - val_accuracy: 0.8597\n",
            "Epoch 20/40\n",
            "101/101 [==============================] - 3s 27ms/step - loss: 0.1971 - accuracy: 0.9301 - val_loss: 0.3820 - val_accuracy: 0.8624\n",
            "Epoch 21/40\n",
            "101/101 [==============================] - 3s 26ms/step - loss: 0.1880 - accuracy: 0.9331 - val_loss: 0.3846 - val_accuracy: 0.8634\n",
            "Epoch 22/40\n",
            "101/101 [==============================] - 2s 23ms/step - loss: 0.1790 - accuracy: 0.9361 - val_loss: 0.3848 - val_accuracy: 0.8641\n",
            "Epoch 23/40\n",
            "101/101 [==============================] - 2s 24ms/step - loss: 0.1699 - accuracy: 0.9409 - val_loss: 0.3891 - val_accuracy: 0.8655\n",
            "Epoch 24/40\n",
            "101/101 [==============================] - 3s 26ms/step - loss: 0.1608 - accuracy: 0.9451 - val_loss: 0.3900 - val_accuracy: 0.8655\n",
            "Epoch 25/40\n",
            "101/101 [==============================] - 3s 28ms/step - loss: 0.1530 - accuracy: 0.9490 - val_loss: 0.3935 - val_accuracy: 0.8665\n",
            "Epoch 26/40\n",
            "101/101 [==============================] - 2s 24ms/step - loss: 0.1458 - accuracy: 0.9528 - val_loss: 0.3957 - val_accuracy: 0.8658\n",
            "Epoch 27/40\n",
            "101/101 [==============================] - 2s 22ms/step - loss: 0.1379 - accuracy: 0.9541 - val_loss: 0.3994 - val_accuracy: 0.8641\n",
            "Epoch 28/40\n",
            "101/101 [==============================] - 2s 22ms/step - loss: 0.1314 - accuracy: 0.9581 - val_loss: 0.4040 - val_accuracy: 0.8648\n",
            "Epoch 29/40\n",
            "101/101 [==============================] - 3s 26ms/step - loss: 0.1242 - accuracy: 0.9607 - val_loss: 0.4091 - val_accuracy: 0.8641\n",
            "Epoch 30/40\n",
            "101/101 [==============================] - 3s 28ms/step - loss: 0.1193 - accuracy: 0.9638 - val_loss: 0.4113 - val_accuracy: 0.8631\n",
            "Epoch 31/40\n",
            "101/101 [==============================] - 2s 22ms/step - loss: 0.1123 - accuracy: 0.9659 - val_loss: 0.4173 - val_accuracy: 0.8614\n",
            "Epoch 32/40\n",
            "101/101 [==============================] - 2s 21ms/step - loss: 0.1071 - accuracy: 0.9679 - val_loss: 0.4198 - val_accuracy: 0.8604\n",
            "Epoch 33/40\n",
            "101/101 [==============================] - 2s 25ms/step - loss: 0.1020 - accuracy: 0.9693 - val_loss: 0.4267 - val_accuracy: 0.8600\n",
            "Epoch 34/40\n",
            "101/101 [==============================] - 2s 23ms/step - loss: 0.0982 - accuracy: 0.9709 - val_loss: 0.4333 - val_accuracy: 0.8614\n",
            "Epoch 35/40\n",
            "101/101 [==============================] - 3s 25ms/step - loss: 0.0931 - accuracy: 0.9734 - val_loss: 0.4366 - val_accuracy: 0.8614\n",
            "Epoch 36/40\n",
            "101/101 [==============================] - 2s 24ms/step - loss: 0.0902 - accuracy: 0.9731 - val_loss: 0.4429 - val_accuracy: 0.8607\n",
            "Epoch 37/40\n",
            "101/101 [==============================] - 2s 21ms/step - loss: 0.0872 - accuracy: 0.9739 - val_loss: 0.4461 - val_accuracy: 0.8621\n",
            "Epoch 38/40\n",
            "101/101 [==============================] - 2s 23ms/step - loss: 0.0817 - accuracy: 0.9765 - val_loss: 0.4547 - val_accuracy: 0.8607\n",
            "Epoch 39/40\n",
            "101/101 [==============================] - 3s 26ms/step - loss: 0.0781 - accuracy: 0.9775 - val_loss: 0.4583 - val_accuracy: 0.8624\n",
            "Epoch 40/40\n",
            "101/101 [==============================] - 3s 26ms/step - loss: 0.0755 - accuracy: 0.9782 - val_loss: 0.4638 - val_accuracy: 0.8610\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.10/dist-packages/keras/src/engine/training.py:3103: UserWarning: You are saving your model as an HDF5 file via `model.save()`. This file format is considered legacy. We recommend using instead the native Keras format, e.g. `model.save('my_model.keras')`.\n",
            "  saving_api.save_model(\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "prediction = model.predict(test_X, batch_size=batch_size, verbose=0)\n",
        "y_pred = prediction.argmax(axis=-1)\n",
        "\n",
        "cf1 = confusion_matrix(test_y, y_pred)\n",
        "cf1"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "nmK0XrekdjzT",
        "outputId": "f5191e64-acf8-42fc-99c1-b6f33323c6df"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[5684,   92,  154],\n",
              "       [ 245,  162,  113],\n",
              "       [ 340,   51,  344]])"
            ]
          },
          "metadata": {},
          "execution_count": 54
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "y_pred[:50]"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "FEDkHQ6be1Sq",
        "outputId": "570d436f-d692-4b42-d58b-16bd3650b6b3"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
              "       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 2,\n",
              "       0, 0, 0, 0, 0, 0])"
            ]
          },
          "metadata": {},
          "execution_count": 33
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "f1_scores_per_class = f1_score(test_y, y_pred, average=None)\n",
        "\n",
        "# In F1 score cho mỗi class\n",
        "for i, score in enumerate(f1_scores_per_class):\n",
        "    print(f\"F1 Score for class {i}: {score}\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YQxluKt965Af",
        "outputId": "28262720-d5fc-4f9f-c494-7f9a6e573436"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "F1 Score for class 0: 0.9318796622673989\n",
            "F1 Score for class 1: 0.39272727272727276\n",
            "F1 Score for class 2: 0.5111441307578009\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "evaluation = f1_score(test_y, y_pred, average='micro')\n",
        "evaluation"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ozVJ5RMEdrU6",
        "outputId": "5f8f4faa-bfde-4944-d41a-e621f0c3b1ac"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.8609603340292276"
            ]
          },
          "metadata": {},
          "execution_count": 34
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "f1_score(test_y, y_pred, average='macro')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UPO34_eqd4i6",
        "outputId": "5caca1ed-b019-4b30-d01a-a1a7a8927615"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.613786100567483"
            ]
          },
          "metadata": {},
          "execution_count": 39
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "accuracy_score(test_y, y_pred)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OM6gsj50d-YR",
        "outputId": "d8e3b463-7d16-4bdc-a4e2-aca57d324054"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.8609603340292276"
            ]
          },
          "metadata": {},
          "execution_count": 36
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "df_cm1 = pd.DataFrame(cf1, index = [\"normal\",\"hate\",\"harassment\"],\n",
        "                  columns = [\"normal\",\"hate\",\"harassment\"])\n",
        "plt.clf()\n",
        "sn.heatmap(df_cm1, annot=True, cmap=\"Greys\",fmt='g', cbar=True, annot_kws={\"size\": 30})"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 447
        },
        "id": "AamEjCjmeeCS",
        "outputId": "1764ddfa-91ef-4407-cc36-337025526b95"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<Axes: >"
            ]
          },
          "metadata": {},
          "execution_count": 37
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 2 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAhAAAAGdCAYAAABDxkoSAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAABz20lEQVR4nO3dd1QU198G8IeOlAVEaSIq9gIqahR7x1hii4nGoLElUdTYYu9GMRpNNPZYUBNrLL/YC7ZEiQUL2BtioWkUEKQJ+/7BYV5mFtidZYEFn885e5KZnZl7Zx1mvnOrgVKpVIKIiIhIBsOizgAREREVPwwgiIiISDYGEERERCQbAwgiIiKSjQEEERERycYAgoiIiGRjAEFERESyMYAgIiIi2RhAEBERkWzGRZ2BLAYGBkWdBdIjHCCVssvIyCjqLJCeMTQs2PdfXT6TSur9TG8CCCIiIn3Bl1r1WIVBREREsrEEgoiISIIlEOoxgCAiIpJgAKEeAwgiIiKJgm6kWRLwFyIiIiLZWAJBREQkwSoM9RhAEBERSTCAUI9VGERERCQbSyCIiIgkWAKhHgMIIiIiCQYQ6rEKg4iIiGRjCQQREZEESyDUYwBBREQkwYGk1OMvRERERLKxBIKIiEiCVRjqMYAgIiKSYAChHgMIIiIiCQYQ6rENBBEREcnGEggiIiIJlkCoxwCCiIhIggGEeqzCICIiItlYAkFERCTBgaTUYwBBREQkwSoM9RhiERERkWwsgSAiIpJgCYR6DCCIiIgkGECoxyoMIiIiko0lEERERBIsgVCPAQQREZEEAwj1GEAQERFJcBwI9fgLERERkWwsgSAiIpJgFYZ6DCCIiIgkGECoxyoMIiIiko0lEERERBIsgVCPAQQREZEEAwj1WIVBREREsjGAICIikjAwMNDZR47Zs2er7F+jRg3h++TkZPj5+cHe3h5WVlbo3bs3oqOjRcd4+vQpunTpAgsLCzg4OOD777/H+/fvRducOXMGXl5eMDMzQ5UqVRAQECD7N2IAQUREJGFoaKizj1y1a9dGZGSk8Pnnn3+E78aOHYsDBw5g9+7dOHv2LCIiItCrVy/h+/T0dHTp0gWpqam4cOECNm/ejICAAMycOVPYJiwsDF26dEGbNm1w/fp1jBkzBkOHDsWxY8dk5dNAqVQqNdlw+fLlGh909OjRsjIBlJz6JgsLCzRu3BjVqlWDnZ0dACAuLg5hYWG4efMmnj9/rrO0qlWrhrp168LV1RUWFhZISkpCdHQ07t+/jxs3biA1NVUn6RTmOWXR8LIscVJSUnD16lU8fvwY8fHxMDExgZOTE+rWrYvy5cvrNK2nT58iJCQEUVFRSEtLg0KhgLu7u/BWok8yMjKKOgtF6t27dwgJCcGTJ08QHx8PALCysoKrqyuqVq0KZ2fnIs5h4SvokSLd3d11dqzHjx9rvO3s2bOxf/9+XL9+XeW7uLg4lC1bFtu2bcOnn34KALh79y5q1qyJoKAgNGnSBEeOHEHXrl0REREBR0dHAMCaNWswadIkvHz5Eqamppg0aRIOHTqEmzdvCsfu27cvYmNjcfToUY3zqnEjyp9//lmj7QwMDLQKIHRh1qxZmD17ttb7BwQEYNCgQVrt27x5c4wbNw6dO3fO8+b7/PlzHDt2DEuWLMGdO3dkp2NlZYVRo0Zh6NCheV7gKSkpuHTpEv78809ZwV92hXVO+iw6OhohISG4ceMGQkJCcPPmTSQmJgrflytXDqdOncp3Oq9fv8aKFSuwb98+vHv3LsdtateujREjRqB9+/b5SuvkyZNYtWoVbt26leP3FhYW6NWrF/z8/FC6dOl8pVXSZF0PISEhCA0NVbkeXFxcEBgYqLP0rly5goCAAJw7dw5paWm5bufo6IhmzZph0KBBqFKlik7SHj9+PA4fPixap+vz02dF+VL74MEDuLi4wNzcHN7e3vD394ebmxuCg4ORlpYmugfUqFEDbm5uQgARFBQEDw8PIXgAAB8fHwwfPhy3bt1C/fr1ERQUpHIf8fHxwZgxY2TlU+MAIiwsTNaBPxQKhQKrVq1C//79Ndre1dUVQ4YMweXLl2U/bLt06YL169fDyclJ7bZmZmZo0aIFqlatKjuAKMxz0kfBwcHYtGkTbty4gZiYmAJP7+LFi/juu+/w5s2bPLe7desW/Pz80KNHD8ybNw+mpqay0klNTcXUqVNx4MCBPLd79+4dfv/9dxw+fBjLly9Ho0aNZKVT0ly9ehWbNm1CSEhIoVwPAPD27VvMnTsXBw8e1Gj76Oho7N27Fx4eHjoJIE6fPq0SPHxodBlApKSkICUlRbTOzMwsxxezxo0bIyAgANWrV0dkZCTmzJmDFi1a4ObNm4iKioKpqSlsbW1F+zg6OiIqKgoAEBUVJQoesr7P+i6vbeLj45GUlIRSpUppdF7sxpkPLi4uOH78OGrXrq3yXVRUFKKjo5GSkgI7OztUrFgRJiYmWqc1ZswYLFmyRKXYLikpCREREXj16hVKlSoFZ2dnlC1bVut0CvOc9FVoaChOnDhRKGlduXIFX3/9NZKTk0XrFQoFXF1dERcXh6ioKKSnpwvf7d+/H+/evcPy5cs1vsllZGRgzJgxKm+PRkZGcHZ2hrW1NZ4/f463b98K371+/RrDhg3Dpk2bUL9+/XycZfEWGhqKkydPFlp60dHRGDJkCB49eqTyXZkyZWBvbw9TU1PExcUhIiJCpXFcfr19+xZz5szR6TGLI10GEP7+/iq/aW4l5h9//LHw/56enmjcuDEqVKiAXbt2afxgLyxaBxDPnz/HX3/9hadPn6rUtS9dujTfGdOF8ePH48aNGxpvHxERofG21tbWOHr0qOhB+/r1ayxcuBC7d+/GkydPRNubmJigfv366Ny5s8Zv9lkGDx6sUoWU9XZ4+vRpld/f2dkZbdu2RY8ePfDRRx/p5TkVVxYWFrlWMcgVFxeHsWPHioKHcuXKYerUqWjXrp1wA4uKisKqVauwc+dOYbvjx4/LqnJbv369SvDQt29fjBgxQngTycjIQGBgIBYsWCD8LSQlJWHMmDE4ePAgrK2t83W+JZEurwcASEhIwLBhw0TBg42NDYYOHYpOnTrB1dVVtH1qairu3LmDc+fOaVxaoc7ixYuFVv26Pr8P1ZQpUzBu3DjROk3bGdna2qJatWp4+PAhOnTogNTUVMTGxopKIaKjo4WSaScnJ1y6dEl0jKx/z+zbSHtuREdHQ6FQyApStAogAgMD8cknn8Dd3R13795FnTp18OTJEyiVSnh5eWlzyAIRHByMs2fPFsixly5dCg8PD2H5zJkz6N27N16/fp3j9mlpabh06RIuXbqE2bNnw8bGRqN0KleujBUrVgjLqampGDhwIHbs2JHrPpGRkfjjjz/wxx9/qBR15aWwzqm4sLS0RO3ateHh4QFPT094eHjg+fPnGDBggE6Ov2HDBlGRuKurK7Zt26ZStOjk5IS5c+fCxcVFFEiuXLkSvXr1Uvu7v3nzBmvWrBGtGz9+PL7++mvROkNDQ3To0AGenp7o168fXrx4ASAzgNm0aVORtW3SF1nXQ506deDp6Yk6dergxYsXGDhwoM7SWLhwIR48eCAsN2rUCMuWLRMaL0uZmpqibt26qFu3LkaOHCkqQdJGVrspIPN6GDFiBH766ad8HbO40mUJRG7VFZpISEjAo0eP4OvriwYNGsDExASBgYHo3bs3AODevXt4+vQpvL29AQDe3t6YP38+YmJi4ODgAAA4ceIEFAoFatWqJWwjraI6ceKEcAxNaRVATJkyBRMmTMCcOXNgbW2NPXv2wMHBAf3790enTp20OWSx0qpVKwwePFhYvnTpEjp37oykpCSNjxEXF6fRduvWrRNFhP379xf+wDURGxur0XaFeU76rm3btmjevDnc3d1Vqox01ePk9evX2Lp1q2jdDz/8oBI8ZPfNN9/gn3/+weXLlwFkFjVv3LgRY8eOzTOt9evXixr6NWrUCMOGDct1e0dHR8yfPx9fffWVsC4gIAC+vr65PshKsjZt2qBZs2Y5Xg9ZQZYuXLp0CXv37hWWPTw8sHbtWo3fCA0MDKBQKLROPzk5GTNmzBB6QPXv31/0QvGhKeheHrmZMGECunXrhgoVKiAiIgKzZs2CkZER+vXrBxsbGwwZMgTjxo1D6dKloVAoMGrUKHh7e6NJkyYAgI4dO6JWrVrw9fXFokWLEBUVhenTp8PPz08IYr799lusWLECEydOxODBg3Hq1Cns2rULhw4dkpVXrX6hO3fuCG9hxsbGSEpKgpWVFebOnYsff/xRm0MWK0uXLhUurrS0NAwdOlTWg1ZTn3zyCdq2bSss79q1S1bwIEdhnVNx4ObmhipVqhToDeTQoUOiouFGjRqpjf4NDAzg5+cnWrdnz548u7xmZGSIHkoAMHLkSLVvV97e3mjYsKGwnJiYiCNHjuS5T0lVGNcDkFn6kPVvaWxsjHnz5hVqnffy5cvx9OlTAJnVoN99912hpU3/7/nz5+jXrx+qV6+Ozz77DPb29vj333+Ftm0///wzunbtit69e6Nly5ZwcnIS/Y0bGRnh4MGDMDIygre3N7788ksMGDAAc+fOFbapVKkSDh06hBMnTqBu3bpYsmQJ1q9fDx8fH1l51aoEwtLSUqh3d3Z2xqNHj4R681evXmlzyGKjYcOGomqa/fv3IzQ0tEDSkhYxF1TDpsI8J8okbY+Q1adbnSZNmsDV1VUoCXn58iWuX7+eayPHq1eviqqgypcvj8aNG2uU1qeffoorV66I8vzFF19otC/JExoaKurB1K5dO1SvXr1Q09+yZYuwPGPGDFhaWhZa+vqoqLpx5lU9DQDm5uZYuXIlVq5cmes2FSpUUNuLpnXr1rh27ZpWecyiVUjdpEkTYWSszp07Y/z48Zg/fz4GDx4sFKOUVEOGDBEt//777wWSjouLiygavHbtGm7fvl0gaRXWOVGmxMRE0YMZAJo1a6bRvgYGBmjatKlo3ZkzZ3LdXtoGqGnTphrfGKXpXLx4kQ3qCsiePXtEy926dSu0tNPS0jB9+nShp0+nTp3Qpk2bQktfXxXVUNbFiVYBxNKlS4W3mDlz5qBdu3bYuXMnKlasiA0bNug0g/qmS5cuouW8bt750alTJxgb/38B0enTpwskHaDwzokyPXz4UDQokKurq6yut9KGynmNvSH9Tk4jZ0dHR5QrV05YTktLw8OHDzXenzQnDfTk9J7Kr3Xr1uH+/fsAMrsPT506tdDSpuJNqyqM7CMgWlpaqrTw1iempqZwd3eHvb090tLS8N9//yEiIkKr+n1HR0fRcMJhYWHCsLKWlpbo06cP+vXrh2rVqsHZ2Rnx8fGIiorC+fPnceDAAVkDs0gH78neHbVevXoYNGgQWrVqBTc3N5iZmSEmJgYPHjzAsWPHsH37do27pBbmOVEmaf9+uQP/VK5cOc/jZScdQle6rzpVqlQRNRR8/PgxPD09ZR2D8vby5UthgB8gsytvVpfZxMREHDt2DIcOHcKTJ0/w8uVLWFlZoUyZMvDy8kLr1q3RqlUrrd9yHz58iLVr1wrL48ePz9c4MiVJUTWiLE7yPZBUQkKCyjj1+WkJrEsrV66Eu7u7SkOktLQ0BAcH48iRI1i1apXG7TakD/Wsm3O7du2wefNm0dsaAJQtWxZly5aFh4cHvv32W1y5cgUjRowQWtHLTcvS0hLLli1TqXIAgIoVK6JixYro0KED5s6di19++QWzZs1SO8hMYZ4TZZKO6qrJyKLZSec9iIiIQEpKiko3seTkZJVAUu6cCdK8cURa3cs+HwEAIaC/cOECpkyZojL65Zs3b/DmzRs8ePAAO3fuRO3atTFz5kzZgV1GRgamT58ulIY1aNAAffr0yceZlCwluepBV7QKsbJm8rK0tISNjQ3s7OxgZ2cHW1tbvermVbt27RxbMZuYmKBJkyaYM2cOwsPDMWfOHI2iTembYnx8PPr374/jx4+rPGhz0rBhQ5w9exY9evSQnVZGRgbOnTuXY/AgZWFhgalTp+Lw4cOwsrKSlU5BnhNl+u+//0TLcgOIMmXKiKq3MjIycuyu++bNG1EPDRMTE9jb28tKS9qtVJp3yr/w8HDRsqWlJf766y8MHTpUo6Gzb926hQEDBsgePXXr1q1CyaaJiQnmzp3LhybJolUJxJdffgmlUomNGzfC0dGxWF90FhYWmDlzJlq0aIFu3bqJ+stLSQdlqlq1KtavXy8EH8+ePcPvv/+O69evIzExEa6urujatSu6du0q7FOqVCns2LEDTZs2xdWrV3NMx8DAQGXUv+XLlwv11xkZGTh48CAOHz6M58+fw9LSEvXr14evr6/ood+hQwcEBATk2cK/sM6J/p+0IaLcrnoGBgYwMzMTlS7ldN1K0zE3N5f9t2phYZHnMSn/pIM/hYeHi8ZjcHJyQrdu3VCjRg2UKlUK0dHROHPmjKjdREpKCsaPH4/t27fnOAy91PPnz0Vz5Hz99dc6nX2yJGAVhnpaBRA3btxAcHBwoXYz0lRGRgaCgoJw6NAhXLp0CXfu3MHr16+RkZEBe3t7eHl5oWvXrhg4cKDoxt2mTRvs2LED3bt3z3XqYOnDtk6dOsL/b9q0CX5+fiptK9auXYvWrVtj7969QumMmZkZtm/fjlq1aonmOMhiY2OjcvE2aNAAQGY32Z49e4rmhwcyx4j44YcfsHbtWtGw0r1794avr6/KoEWFfU70/6QPYW1GqDM3NxcFDTk92KVBhTbpSPdhAKF70gAie0PVnj17YsaMGSpBZt++fXHx4kWMHj1aaLOUlpaG8ePH4+DBg6ISqpzMnDlT+Ld0d3dX6TJOrMLQhFYhVqNGjfDs2TNd5yXfjh8/jho1aqB58+bw9/dHYGAgIiIikJycjNTUVERGRuLQoUMYPnw4qlatqvIQ7tq1K0aMGJHr8XOrDvjrr78wePDgXBtmnjlzBt27dxc9WKtVq4bPPvtMVjrv379Hly5dVPKdJTExEb6+vjh27JhofV6tqgvrnOj/SWfl02ZCMulMnNLJuACozJFSUOlQ/uRW6tmmTRvMnz8/1xKqxo0bY8WKFaKXjfDwcBw9ejTP9P78808EBQUByHxIzpkzR/bMrh8CduNUT6sAYv369fjxxx+xefNmBAcHIyQkRPQpKkFBQaJx5PPy4sULtG/fHhcuXBCtnz59eq5/sDndPNPS0lRGB8zJ33//LRqoBVAdKCqvdIDM3106SYqUUqnE8OHDRQ/2GjVqoFWrVhqnVRDnRP9P+lafvUunpqTBQU6lC9KHQkGlQ/mT029qbGyMGTNmqH34NGrUCN27dxet27VrV67bx8TEYPHixcLyp59+KhpxlEgOrQKIly9f4tGjRxg0aBAaNWqEevXqoX79+sJ/i4uUlBQMGDBAdGN1dHREx44dc9w+ISFBZd3Ro0c1nh9h3bp1ouXGjRvn+FaYUzoA8Ntvv2mUTlhYmMr0w7kFEIV1TvT/pO0KpCUSmpAGftJjAlAZSVCbdKT75JQO5U9Ov2nz5s017jEj7TkREhKiEvhlmTdvnlDlUaZMGUyYMEFmbj8choaGOvuUVFqd2eDBg1G/fn0EBQXh8ePHCAsLE/1XnZSUFMTHx4s+ReXRo0f466+/ROvkBBByZvu8fPmyqA65VKlSOU5Wk5ycrNL9Mj4+Xtawo9J85faWUVjnRP9P+sCQOyaJUqlUebDnNOywNJ3k5OQ8583IibTNAwMI3cvpN5V2r86Lh4cHzM3NheWUlJQcS2KPHj0qerGYOnWq3nS510eswlBPqwAiPDwcP/74Ixo3boyKFSuiQoUKoo86/v7+sLGxEX2KknRegtwah0rnTwcgjOCmifT0dJVBf7KmW5WSdt96+PChrJv/vXv3NEqnMM+JMkm7UmYfREgTr169EgWYhoaGOXaftrOzE928sgZSk0N6fcjtBkrqlSlTRmVdxYoVNd7f2NhYNBgckHN32+xVF61atcLHH3+seSaJcqBVL4y2bdvixo0bskfQyzJlyhSMGzdOtK4ogwhpg9DcRmLLachguaUn0u1zGzfjzp07cHFxKZR01O2rq7QoU6VKlUTLkZGRsvaXbu/i4pJjPbq5uTlcXFxEI0lGRETk+MDSNC129dO9nH5TdeO3qNs+p7/h7L09zp49i5o1a8pKA8i8fqT77d27V6tj6buSXPWgK1oFEN26dcPYsWMRGhoKDw8PlTrvTz75JM/9zczM9KoxlrRxWW51+DlNZiX3PLIXNQK5d4u7ffs22rVrVyjpSBVUWpRJ+sCQO7+EtMQnr4d6pUqVRAHEo0ePZI1YKCct0k5Ow4vn1oYhN9IqLenfJMlXkqsedEWrAOLbb78FANH84lkMDAyK3TgA0pEAX758meN2r1+/Rnh4uKiaRjpSnzrS4v3cipSlAzIVVDqFeU6UqUqVKjAxMREC1xcvXiAmJkbjqh/ptZHX21/NmjVF3X6vXbuGnj17apROTEyMKPgwMTGRPZcGqWdnZwdnZ2dRaY+mw+tnyT5lO6A6vgtRQdAqgMhtoKXiqnnz5qLlvMa4+OuvvzBq1ChhuUGDBrkO0iTl4OCgUleZW3uDQ4cOIT09HUZGRgAy3/zs7Ozw5s0bjdLKGngqi7RNRHaFdU6UycrKCg0bNhT64gOZ8x5oMhy4UqlU6Xqc19TLrVu3FvXeuXDhApRKpUZvV+fPnxctN27cOMfGmpR/bdu2xR9//CEs3759W+Ph4V+9eqXSjianNhQrVqxQOzeO1L1797Bo0SJhuUyZMvjxxx9F27i5uck6ZnHBKgz1ZP9CaWlpMDY2VpkApriysbFB7969ReukjSqz27dvn2i5R48eGhd1SYeUvn37dq5j3b98+VLlBt6rVy+N0jEyMlJ5y8xriu7COif6f23bthUt//nnnxrt9++//4q62JYpUwZ169bNdXsvLy9Rm5Rnz57h4sWLGqUlzZM0z6Q77du3Fy2fPHlS4xc16cBxlStXzrGdy0cffYSmTZvK+kiHxTY1NVXZpqQGleyFoZ7sAMLExARubm7FrpoiNz/99JPoBpuSkoIjR47kuv3Zs2dx9+5dYblChQoYOHCg2nRKlSqF8ePHi9blNeALANE0uwDw/fffazRi3LBhw0R9yOPi4lRuMtkV5jlRpi5duoi6712+fFlUIpETpVKJlStXitb16tUrzzclQ0NDlWBy5cqVanv0BAUF4cqVK8KypaUlW+0XoEaNGoka10ZGRmL//v1q90tKSkJAQIBoXadOnXScO6KcaVVGM23aNEydOlWl3q0oTZo0SZhsShNGRkb46aefMHToUNH6NWvW5NmtLmsK3Ox+/vnnPAfQMjQ0xIYNG0QN0BISEvDrr7/mmcft27eLRvasXr061q5dm2dE+9FHH4mKHAFg1apVefasKMxzokz29vaiOUuAzFFQc+pWm2Xt2rWiadOtra01mp112LBhomDl0qVLeQ5KFh0djWnTponWDRgwAKVLl1abFmnHyMgI3333nWjdwoULcevWrVz3SU9Px/Tp00UlUhYWFirXFWmHJRDqadUGYsWKFXj48CFcXFxQoUIFlSKsopiRsVOnTli4cCHOnz+PXbt2ITAwEHfv3lUpKVEoFOjcuTMmTpyo8oB8+PBhjg1Dpfbs2YMzZ86gdevWADIbLJ0+fRozZ87Epk2bRN2lGjRogMWLF6vUU48bN05tAKZUKjF27FicOHFCeMv86quvUK5cOUyePFn0OysUCgwZMgRz5swRzeR57949LFiwQG/OqbgIDg7OceTG7CU1QGaJlbRNQhYHB4c8uzoPHToU+/fvFxrtPn/+HP369cO0adPQtm1b4cYTFRWFVatWYefOnaL9v/32W40ay5UuXRrffvstli5dKqxbsmQJIiIiMHz4cKHRbEZGBk6dOoX58+cjIiJCdB6DBw9Wm05JdvXq1RyHfZe2LUpNTdX6evDx8UGjRo2EIPHt27f46quvMGrUKPTq1UvUVfPmzZtYvHixytD2kyZNYjdqHWEbCPUMlHKHpgMwZ86cPL+fNWuW/IzkM0o7ffq08PDLkpycjOfPnyMuLg7p6emwt7dHxYoVhYaJ2UVGRqJly5Yad6lzcnLC+fPnVbq1paSk4NGjR8LU1zkNR7tu3Tp88803Gp/bxIkTVRouZeU5azrvypUrq3S/fPXqFdq0aaNxe5XCPCd1tLgsdapt27aiHgja6NmzJxYuXJjnNpcvX8aQIUNUghWFQgFXV1fEx8cjMjJSJRBu164dVq5cqfHfTUZGBkaMGIHTp0+L1hsZGcHFxQXW1tZ4/vy5SkmVubk5Nm7cqNIot7AVdcPtdu3aiYIqbfTo0QP+/v55bhMTE4P+/furDCWfVXVcqlQpREVF5dhLo0+fPhq9AMlx6dIlUXWmi4tLnm3EClNBP+Bbtmyps2OdO3dOZ8fSJ1qVQGgTIBQFc3NzjQa7OnToEAYNGpRr982cREVFoWXLlti3b59o2FkzMzPUqlUrx33S09Mxd+5c2X/kixYtwrt377BkyRJRGwhnZ+dcx8u/e/cuunXrJmuMgcI8J8rUqFEjrF27FmPGjEFsbKywPj4+PscxOoDMWWP9/f1lBd2GhoZYtmwZpkyZgkOHDgnr09PTc+11ZGtri+XLlxd58PAhcXBwwNatWzFq1ChR4J+WlqYyJkcWQ0NDjBgxQqMJ8Ih0KV8hXHBwMH7//Xf8/vvvsuZpKAjz58/H6tWrcfPmTY26Kr19+xa7du1Cy5Yt0bVrV1nBQ5YXL17A29sbfn5+eT6ok5OTsXfvXtStW1frB+2KFSvg6emJHTt25DnIzOPHjzF69Gh4enrKHqAIKNxzokze3t44dOgQ+vXrl+tMsABQq1Yt/PrrryqBpKbMzMywdOlSLF++PM+xIywsLPDFF1/g8OHDaNy4sex0KH+cnJywfft2TJ8+Pc8ukqampmjfvj327dvH4KEAcDIt9bSqwoiJiUHfvn1x5swZoQ42NjYWbdq0wY4dO3IdCjrPjOiwoUmpUqVQq1YtVKxYEc7OzrCysoKhoSFiY2Px5s0b3L59G6GhoTovFq1duzbq168PFxcXGBoa4tWrVwgPD8c///wje8KkvFhbW6Np06aoWrUqbGxskJCQgOjoaFy9elXnYzAU1jlJFXUVRlFJTk7GtWvX8OjRI8THx8PExASOjo6oW7euRvPMyBEeHo4bN24gOjoaaWlpUCgUqFy5Mry8vPRqpFig6KswitL9+/dx584dxMTEQKlUwtbWFuXKlYOXl1eeAWdJV9AP5rzGV5FLWnVYUmgVQHz++ed4/PgxtmzZIrzJ3L59GwMHDkSVKlWwfft2+RkpwS1VSb4PNYCgnH3IAQTljAFE0dOqDUTWtLDZi0Fr1aqFlStX5joVNhERUXHBl1r1tB7KOqcJp0xMTPimQERExV5JbrugK1r9Qm3btsV3330n6tb04sULjB07VjSDJBEREZVMWgUQK1asQHx8PCpWrIjKlSujcuXKqFixIuLj4zkSIRERFXsciVI9raowypcvj6tXryIwMBB37twBkDltsHRCGCIiouKIVRjqaRVAAMCpU6dw6tQpxMTEICMjA9euXcO2bdsAABs3btRZBomIiEj/aBVAzJkzB3PnzkXDhg3h7OxcootoiIjow8PnmnpaBRBr1qxBQEAAfH19dZ0fIiKiIscAQj2tAojU1FQ0bdpU13khIiLSC2wDoZ5Wv9DQoUOF9g5ERET04dGqBCI5ORnr1q3DyZMn4enpqTKo1NKlS3WSOSIioqLAKgz1tAogQkJCUK9ePQAQTTkL8EcnIqLij1UY6mkVQJTUiUGIiIhIM1qPA0FERFRSsTRdPQYQREREEqzCUI+/EBEREcnGEggiIiIJVmGoxwCCiIhIggGEeqzCICIiItlYAkFERCTBEgj1GEAQERFJMIBQjwEEERGRBAMI9dgGgoiIiGRjCQQREZEESyDUYwBBREQkwQBCPVZhEBERkWwsgSAiIpJgCYR6DCCIiIgkOJmWevyFiIiISDaWQBAREUmwCkM9lkAQERFJGBgY6OyjrYULF8LAwABjxowR1iUnJ8PPzw/29vawsrJC7969ER0dLdrv6dOn6NKlCywsLODg4IDvv/8e79+/F21z5swZeHl5wczMDFWqVEFAQIDs/DGAICIi0jOXL1/G2rVr4enpKVo/duxYHDhwALt378bZs2cRERGBXr16Cd+np6ejS5cuSE1NxYULF7B582YEBARg5syZwjZhYWHo0qUL2rRpg+vXr2PMmDEYOnQojh07JiuPBkqlUpm/09QNFhdRdnpyWZKeyMjIKOoskJ4p6EaOAwYM0NmxtmzZImv7hIQEeHl5YdWqVfjhhx9Qr149/PLLL4iLi0PZsmWxbds2fPrppwCAu3fvombNmggKCkKTJk1w5MgRdO3aFREREXB0dAQArFmzBpMmTcLLly9hamqKSZMm4dChQ7h586aQZt++fREbG4ujR49qnE+WQBAREUkUZRWGn58funTpgvbt24vWBwcHIy0tTbS+Ro0acHNzQ1BQEAAgKCgIHh4eQvAAAD4+PoiPj8etW7eEbaTH9vHxEY6hKTaiJCIiktBlqXhKSgpSUlJE68zMzGBmZqay7Y4dO3D16lVcvnxZ5buoqCiYmprC1tZWtN7R0RFRUVHCNtmDh6zvs77La5v4+HgkJSWhVKlSGp0XSyCIiIgKkL+/P2xsbEQff39/le2ePXuG7777Dn/88QfMzc2LIKfyMIAgIiKS0GUVxpQpUxAXFyf6TJkyRSXN4OBgxMTEwMvLC8bGxjA2NsbZs2exfPlyGBsbw9HREampqYiNjRXtFx0dDScnJwCAk5OTSq+MrGV12ygUCo1LHwAGEERERCp0GUCYmZlBoVCIPjlVX7Rr1w6hoaG4fv268GnYsCH69+8v/L+JiQkCAwOFfe7du4enT5/C29sbAODt7Y3Q0FDExMQI25w4cQIKhQK1atUStsl+jKxtso6hKbaBICIi0gPW1taoU6eOaJ2lpSXs7e2F9UOGDMG4ceNQunRpKBQKjBo1Ct7e3mjSpAkAoGPHjqhVqxZ8fX2xaNEiREVFYfr06fDz8xOClm+//RYrVqzAxIkTMXjwYJw6dQq7du3CoUOHZOWXAQQREZGEvg4t8PPPP8PQ0BC9e/dGSkoKfHx8sGrVKuF7IyMjHDx4EMOHD4e3tzcsLS0xcOBAzJ07V9imUqVKOHToEMaOHYtly5bB1dUV69evh4+Pj6y8cBwI0kt6clmSnuA4ECRV0ONADB06VGfHWr9+vc6OpU/YBoKIiIhkYxUGERGRBEvF1WMAQUREJMEAQj1WYRAREZFsLIEgIiKSYAmEegwgiIiIJBhAqMcAgoiISIIBhHp6E0CkpaUVdRZIj7DfP2XHmzmR/tGbAIKIiEhfMGhVjwEEERGRBAMI9diNk4iIiGRjCQQREZEESyDUYwBBREQkwQBCPVZhEBERkWwsgSAiIpJgCYR6DCCIiIgkGECoxyoMIiIiko0lEERERBIsgVCPAQQREZEEAwj1GEAQERFJMIBQj20giIiISDaWQBAREUmwBEI9BhBEREQSDCDUYxUGERERycYSCCIiIglDQ75fq8MAgoiISIJVGOoxxCIiIiLZWAJBREQkwRII9RhAEBERSTCAUI8BBBERkQQDCPXYBoKIiIhkYwkEERGRBEsg1GMAQUREJMEAQj1WYRAREZFsLIEgIiKSYAmEegwgiIiIJBhAqMcqDCIiIpKNJRBEREQSLIFQjwEEERGRBAMI9ViFQURERLKxBIKIiEjC0JDv1+owgCAiIpJgFYZ6DCCIiIgkGECoxzIaIiIiko0lEERERBIsgVCPAQQREZEEAwj1WIVBREREsrEEgoiISIIlEOoxgMhGqVTixYsXuH//PqKjo/H27VuYmppCoVCgQoUKqFOnDszMzIo6m/QBePfuHUJCQvDkyRPEx8cDAKysrODq6oqqVavC2dm5iHNIJc3Tp08REhKC6OhopKWlQaFQwN3dHfXr1/8g73sMINT74AOIuLg4nDp1Cv/88w8uXryIN2/e5LqtsbExWrVqBV9fXzRq1KhA8vPPP//gm2++UVm/adMmfPTRRxodY+rUqfjf//6ndR5GjBgBPz8/rfcvSaKjoxESEoKQkBCEhobi5s2bSExMFL53cXFBYGCgztK7cuUKAgICcO7cOaSlpeW6naOjI5o1a4ZBgwahSpUqeR5zxYoVWLlypdZ56tGjB/z9/bXevyTJfj2EhITkeD2cOnUqX2mkpqbi7t27uHHjBkJDQxESEoLw8HAolUphmwULFqBXr175SifLyZMnsWrVKty+fTvH7y0sLNCzZ0+MHDkSdnZ2OkmTSoYPOoCYN28e9uzZk+eNOrv3798jMDAQgYGB+OSTTzBt2jRYWVnpLD+JiYmYPXu2zo5H2rl69So2bdqEkJAQxMTEFEqab9++xdy5c3Hw4EGNto+OjsbevXvh4eGhNoCg/Mm6Hm7cuFGg18OPP/6IK1eu4O7duxrfk/IjNTUV06ZNw4EDB/Lc7t27d/jjjz9w5MgRLFu2rMBenvQNSyDU+6ADiNDQ0Bz/UI2MjFC2bFnY29sjLS0NkZGRePv2rWibv/76C2FhYdiwYQMsLS11kp+ff/4ZkZGROjkWaS80NBQnT54stPSio6MxZMgQPHr0SOW7MmXKwN7eHqampoiLi0NERATev39faHmjzOvhxIkTBZ7On3/+qXKfKSgZGRkYO3asSumZkZERnJ2dYW1tjefPn4vy8/r1a3z99dfYuHEj6tevXyj5LEoMINT7oAOI7BQKBTp37oxWrVqhQYMGoqAgPT0dwcHBWLFiBYKDg4X1oaGhmDZtGn755Zd8p3/16lXs2LEDQOYY7CYmJkhJScn3cYHMNxt7e3uNt3d1ddVJuiWRhYUF3r17p7PjJSQkYNiwYaLgwcbGBkOHDkWnTp1U/i1SU1Nx584dnDt3TuPSCqmJEyeievXqGm/v4OCgVTofAl1fDzkxNjaGoaEhUlNTdXbMDRs2qAQPffv2xfDhw+Ho6AggM8gIDAyEv78/IiIiAABJSUkYO3YsDhw4AGtra53lh4qnDz6AKFeuHL755ht06dIF5ubmOW5jZGSEjz76CJs2bcK8efOwe/du4bsTJ07g4sWLaNy4sdZ5SE1NxcyZM4U6zn79+uH06dPCH21+1a9fH+XKldPJsT4klpaWqF27NurUqQNPT0/UqVMHL168wMCBA3WWxsKFC/HgwQNhuVGjRli2bFmudc2mpqaoW7cu6tati5EjR2r1xlq7dm2N29PQ/8u6Hjw8PITP8+fPdXo9GBgYwM3NDZ6envDw8ECdOnVQu3ZtDB06FJcvX9ZJGm/evMGaNWtE68aNG4evv/5atM7Q0BAdOnSAp6cnvvjiC7x48QIAEBUVhU2bNmH06NE6yY++YgmEeh90AOHn5wdvb2+YmppqtL2RkRFmzJiB27dv49atW8L6PXv25CuAWLVqFcLCwgAATk5O+O6773D69Gmtj0f506ZNGzRr1gzu7u4qM/Jl3UR14dKlS9i7d6+w7OHhgbVr16JUqVIa7W9gYACFQqGz/FDO8roenj9/rrN0Vq1aherVqxf4v+mGDRtEDT8bNmyIYcOG5bq9o6MjfvjhBwwaNEhYt3nzZvj6+pboRpWcjVO9D/oXatWqlcbBQxYjIyMMGTJEtO78+fNa5+HOnTvYtGmTsDxt2jSdtakg7bi5uaFKlSoFfgNZuHChUOpkbGyMefPmaRw8UOEprOuhUaNGBR48ZGRkiIJWABg5cqTat21vb280bNhQWE5MTMTRo0cLJI/6wsDAQGcfOVavXg1PT08oFAooFAp4e3vjyJEjwvfJycnw8/ODvb09rKys0Lt3b0RHR4uO8fTpU3Tp0gUWFhZwcHDA999/r9J26syZM/Dy8oKZmRmqVKmCgIAA2b+R1n8Rjx49wvTp09GvXz+hZfKRI0dEb+YllZeXl2g5NjYWSUlJso/z/v17zJw5U/iH7dChA9q2bauTPJJ+Cw0NxZ07d4Tldu3ayWqXQKSNa9eu4fXr18Jy+fLlNS497d27t2i5MBsaf0hcXV2xcOFCBAcH48qVK2jbti26d+8uPFuz2qDs3r0bZ8+eRUREhKhLb3p6Orp06YLU1FRcuHABmzdvRkBAAGbOnClsExYWhi5duqBNmza4fv06xowZg6FDh+LYsWOy8qpVAHH27Fl4eHjg4sWL2Lt3LxISEgAAN27cwKxZs7Q5ZLFiY2Ojsi7rN5Bj06ZNQt9ra2trTJs2Ld95o+Jhz549ouVu3boVUU7oQ3LmzBnRctOmTTV+Q27atKlo+dKlSwXegLQoFVUJRLdu3dC5c2dUrVoV1apVw/z582FlZYV///0XcXFx2LBhA5YuXYq2bduiQYMG2LRpEy5cuIB///0XAHD8+HHcvn0bv//+O+rVq4ePP/4Y8+bNw8qVK4WGuGvWrEGlSpWwZMkS1KxZEyNHjsSnn36Kn3/+WVZetQogJk+ejB9++AEnTpwQVQG0bdtWOImSTFpcBOQcVOTlyZMnWL16tbA8duxYlC1bNt95o+Lh7NmzomU2aqTCcPfuXdGynO6Yjo6OosbYaWlpOXY9LimKKoDILj09HTt27EBiYiK8vb0RHByMtLQ0tG/fXtimRo0acHNzQ1BQEAAgKCgIHh4eQm8aAPDx8UF8fLxQihEUFCQ6RtY2WcfQlFaNKENDQ7Ft2zaV9Q4ODnj16pU2hyxWsnflBDJHn5PTlkKpVGLmzJlCN8369evjs88+02kepd69e4eIiAjExcXBzMwMNjY2cHR0lN0GhPLv5cuXiIqKEpbLlSsndIlLTEzEsWPHcOjQITx58gQvX76ElZUVypQpAy8vL7Ru3RqtWrXKdwvx1NRUPHv2DLGxsTA2NoatrS0cHBzYBqOEkz7wK1euLGv/KlWqiBoSP3r0CB4eHjrJW0mWkpKi0i3fzMws1yHCQ0ND4e3tjeTkZFhZWWHfvn2oVasWrl+/DlNTU9ja2oq2d3R0FO4pUVFRouAh6/us7/LaJj4+HklJSRrfB7QKIGxtbREZGYlKlSqJ1l+7du2D6C64b98+0XKLFi1k7b9jxw4hCDExMcHs2bMLtMvQyJEj8fjxY5VGNGZmZvD09ETr1q3Rp08fNt4sJDdv3hQtly9fHgBw4cIFTJkyRWW0wzdv3uDNmzd48OABdu7cidq1a2PmzJnw9PTUKv25c+fi+fPnKjc0Y2Nj1KpVCy1btkS/fv1QunRprY5P+ik5OVlloDq5c6o4OTmJlrN6j5VEurwn+/v7Y86cOaJ1s2bNynXk4erVq+P69euIi4vDn3/+iYEDB6qUWuoDraow+vbti0mTJiEqKgoGBgbIyMjA+fPnMWHCBAwYMEDXedQr586dw5UrV0TrevToofH+kZGRonqmIUOGFPhQxPfv389x9MKUlBRcvnwZixcvRrt27fDHH38UaD4oU3h4uGjZ0tISf/31F4YOHarRUMm3bt3CgAEDtB4d8dGjRzkOUvb+/XuEhIRgxYoVaNeuHZYvX4709HSt0iD98+bNG9F8GiYmJrIGmAOg8taavUFmSaPLKowpU6YgLi5O9JkyZUquaZuamqJKlSpo0KAB/P39UbduXSxbtgxOTk5ITU1FbGysaPvo6GghuHNyclKpZs9aVreNQqGQVQqpVQCxYMEC1KhRA+XLl0dCQoLw1tK0aVNMnz5dm0MWC7GxsSpRZLt27WS9Cc6dO1fog12pUqUcJ84qCm/fvsWCBQswYcIEDpVcwKSDP4WHh2PGjBnCzd3JyQnDhg3DkiVLsGrVKsyaNQutWrUS7ZOSkoLx48cXWK+n5ORkrF69GoMHDxaNGUDFl7TBo7m5uey3bOnDhdeGZszMzIRumVkfOTOcZmRkICUlBQ0aNICJiYloFNF79+7h6dOn8Pb2BpDZ5TY0NFT0MnLixAkoFArUqlVL2EY6EumJEyeEY2hKqyoMU1NT/Pbbb5g5cyZCQ0ORkJCA+vXro2rVqtocrljIyMjA5MmTRXXX1tbWeUaRUgcOHMC5c+cAZEa3s2bNKrA2CMbGxmjUqBFatGiB2rVro2LFilAoFMjIyMDr169x8+ZNnDx5EseOHRMFDEeOHIFCoRB1+SHdkgYQDx8+FP6/Z8+emDFjhsqNum/fvrh48SJGjx4tTO+dlpaG8ePH4+DBgzA2zvtP2cDAAPXq1UOrVq3g4eGBypUrw8bGBoaGhoiNjcWtW7dw9uxZ7N+/X1Q6cenSJYwfPx4rV66EkZFRfk+dipA0gNBmim7paL0lvRdGUZgyZQo+/vhjuLm54e3bt9i2bRvOnDmDY8eOwcbGBkOGDMG4ceNQunRpKBQKjBo1Ct7e3mjSpAkAoGPHjqhVqxZ8fX2xaNEiREVFYfr06fDz8xP+zb/99lusWLECEydOxODBg3Hq1Cns2rULhw4dkpVXrQKIuXPnYsKECShfvrxQfwtkjpO+ePHiEvnw+emnn/D333+L1s2aNUvjOsT//vsPCxcuFJZ79epVYLPade3aFaNHj1apr8zi4uICFxcXdOzYEV9//TXGjRsnaly1c+dOtGjRAm3atCmQ/H3ocntra9OmDebPn5/rjatx48ZYsWIFvvrqK2RkZADILL04evQounbtmmt6zZo1Q5cuXVTaLGVxcHCAg4MD2rRpg2+//Rbjx4/H1atXhe/Pnj2L7du348svv9T0FEkPSautTExMZB9D+sKjq/l69FFRBRAxMTEYMGAAIiMjYWNjA09PTxw7dgwdOnQAkDnpoqGhIXr37o2UlBT4+Phg1apVwv5GRkY4ePAghg8fDm9vb1haWmLgwIGYO3eusE2lSpVw6NAhjB07FsuWLYOrqyvWr18PHx8fWXnVqgpjzpw5OY578O7dO5Ui/pLg999/x+bNm0XrBg8ejI8//ljjYyxYsECot7K3t8f48eN1mUWRpk2b5ho8SFWpUgWbN2+Gm5ubaP3y5ctF9aWkOzm9+RkbG2PGjBlqb1qNGjVC9+7dRet27dqV5z7169fPNXiQcnJywsaNG1GvXj3R+jVr1mg1WBrpD+l1p82U4dIJvdiLS/c2bNiAJ0+eICUlBTExMTh58qQQPACZpUArV67E69evkZiYiL1796rc7ytUqIDDhw/j3bt3ePnyJX766SeVUsrWrVvj2rVrSElJwaNHj/DVV1/JzqtWAYRSqczxRnfjxo0S13L74MGDopIDILPR5Lhx4zQ+xqlTp0TDvk6ePFn2uBEFyc7OTiXwu3//vkqfcdINCwsLlXXNmzfXuDSrT58+ouWQkBCdztRoZmaGhQsXim44//33X76GbKeiJ73utCk9SE5OFi2X5J5b+jAOhL6TFUDY2dmhdOnSMDAwQLVq1VC6dGnhY2Njgw4dOmg0nkFKSgri4+NFH30sCjtz5gymTZsmehNv37495s6dq/FF8fbtW8ybN09YbtGiBTp37qzzvObXRx99JDSwyXLhwoUiyk3JllMAIac6y8PDQ1QXnZKSIprRUxcqVKigUoXFAKJ4k153ycnJsksZpaVQOV3LJQUDCPVktYH45ZdfoFQqMXjwYMyZM0f0Fm1qaoqKFStq1Iozpz6xM2bM0Ku2ExcvXsS4ceNEDQybNm2KxYsXy2pMtmXLFqE1bKlSpfTqHKUaN24sDK0NZI6WSbpXpkwZlXUVK1bUeH9jY2OUL19eFDT8999/usiaSJMmTURdRXk9FG92dnYwMDAQgoa0tDT8999/OV6PuZF2/StpJc7ZcTZO9WQFEFnz3leqVAlNmzbVqhEOkNnKVFoFoE8tvENCQjBy5EhRqUi9evWwbNky2XV+WS3mgczoPXtdlhzZp9IFgC+//FJWDxBNSIvQS3If76Lk7u6uss7KykrWMaTbZ7/OdIXXQ8libm4OZ2dnRERECOsiIyNlBRDSgahyupbpw6FVL4zsfdKTk5NV6l/VTUmb0xCe+jL2wL179/DNN9+IuifVrFkTa9asKdHFdQBUGtnoy79JSZPT8MFy2zBIq/yk3et0gddDyePu7i4KIB4+fChrKGrpUNglOYAoyVUPuqJVGc27d+8wcuRIODg4wNLSEnZ2dqJPcRUWFoZhw4aJ3ubc3d2xbt06Ya6Ckkw6j0lJLp4sSnZ2dipv93LnkJGWBkjHxtcFaZ6K8982ZapZs6Zo+dq1axrvGxMTI5oHw8TERPZcGsUJ20Cop1UJxPfff4/Tp09j9erV8PX1xcqVK/HixQusXbtWpcdCcREREYGhQ4eK6pKz+sbm50Har18/tG7dWvZ+kyZNEuXl+++/R/Xq1YVluWPYayJ7339Addx70p22bduKhg6/ffu2xkOiv3r1SjSgGSCvDYWmpJPGFcQ1R4WrdevW+O2334TloKCgXHvVSUkb0X700UcluhcGqadVAHHgwAFs2bIFrVu3xqBBg9CiRQtUqVIFFSpUwB9//IH+/fvrOp8F6uXLlxg8eLDopuzo6IgNGzaojP0uV8WKFbW6uUureGrVqlWgUz6Hh4erzPHRuHHjAkvvQ9e+fXtRAHHy5ElMnjxZo4Zbx44dEy1XrlxZVj22JuLj41Xm2sga6Y6Kr/r168POzg5v3rwBADx79gwXL17U6N92z549ouV27doVSB71RUkuOdAVraowXr9+LdR9KRQKoTi1efPmwlDNxUVsbCyGDh2KZ8+eCetKly6N3377Da6urkWYs8KTnp6OH374QVTHbWtrCy8vryLMVcnWqFEj0eBOkZGR2L9/v9r9kpKSEBAQIFrXqVMnHecOWLRokagqz8TERPass6R/DA0N0bNnT9G6lStXqu3OGRQUJHrBsLS0lDWQXnFkaGios09JpdWZubu7C9O41qhRQxgJ78CBAwVSF1tQEhMT8c0334jmIlAoFFi3bl2xrdtbtmyZrO52SUlJmDx5ssqYD19//TVHmStARkZG+O6770TrFi5cmOfkWOnp6Zg+fTqeP38urLOwsMizxO+3336TNeHW+/fv8eOPP6q8bfbt2xcODg4aH4f017Bhw0QNwi9fviyq1pCKjo5WmSRxwIABbBND2lVhDBo0CDdu3ECrVq0wefJkdOvWDStWrEBaWhqWLl2q6zwWGD8/P9y8eVO0bsCAAYiNjUVQUJCsY9WqVUsvRpc8ePAg1q9fjyZNmqBjx45o1KgR3NzcVKLg169f4+TJk9i4caOo9AUAPD090a9fv8LMtt65evWqyqh7QGYvnexSU1NzHXDLwcEhz6nafXx80KhRI1y+fBlA5qBjX331FUaNGoVevXqJumrevHkTixcvxqVLl0THmDRpUp438r///htLly5F/fr10alTJzRp0gTu7u4qPSzevn2Lc+fOYcOGDbhz547oOzc3NwwfPjzXND4EuV0P0tFa83M9xMTEiF5mspN203306FGu6TRo0CDPibLs7OzwzTff4OeffxbWLV26FJGRkfj222+FatuMjAycOnUKCxYsEPXccHBwUOlWXhKxCkM9A6UOJjwIDw9HcHAwqlSpImtq6+yKootY7dq1dXasTZs26bSNQocOHUR/tJoeX7ofkDmAlYODg/BAevPmjco2Wdzd3bFly5Yif7so6mK/du3a5fobaapHjx7w9/fPc5uYmBj0799fVKoAZFYZuLm5oVSpUoiKisqxl0afPn1EE+TkZMCAAUKAksXU1BROTk6wsrKCkZERYmNj8eLFC2GCruzKlCmDrVu3FkgjTTmK+mbetm1bnVwPeTUy37t3L6ZOnZqvNIDM9jTqql8zMjIwYsQInDlzRrTeyMgILi4usLa2xvPnz1UCF3Nzc2zcuFEvqjcL+prYtGmTzo5VUgMurUogACAwMBCBgYGIiYlRufFs3Lgx3xkj3UlKSkJ4eLja7bp3745p06axZXUhcnBwwNatWzFq1ChRaVhaWppKn/sshoaGGDFiBPz8/LRKMzU1FU+fPlW7XcuWLbFgwQLY29trlQ7pL0NDQyxbtgxTpkzB4cOHhfXp6ekqJZJZbG1tsWzZMr0IHkg/aBVAzJkzB3PnzkXDhg3h7Oxc5G8H9P8mTJiAc+fO4erVq3j27JnaxlEKhQLt27fHF198odJHnAqHk5MTtm/fjp07d2LLli25PtxNTU3RsmVLjBo1CtWqVdPo2N988w3c3d0RHByMsLAwpKen57m9hYUFWrZsiS+++KLAppsn/WBmZoalS5fCx8cHa9asUam6ymJhYYEePXrAz8/vgwom+VxTT6sqDGdnZyxatAi+vr46ywhHudO9hIQEPHr0CBEREXj16hWSkpJgYGAAa2tr2NjYoFq1anB3d9fLP5SirsIoSvfv38edO3cQExMDpVIJW1tblCtXDl5eXihVqpTWx01KSsKjR4/w4sULvHz5Eu/evUNGRoZwPVSuXBnVqlXTq2Hls+jjNVrShIeHIyQkBNHR0UhLS4O1tTUqV64MLy+vPNtUFJWCviY2b96ss2NlTQNR0mgVQNjb2+PSpUs67anAAIKy+5ADCFLFAIKkCvqa2Lp1q86OpcuXbX2i1V166NCh2LZtm67zQkRERMWExm0gss+emZGRgXXr1uHkyZPw9PRUmZWzOHXlJCIikmKpl3oaBxDSSVfq1asHACrjKPBHJyKi4o7PMvU0DiBOnz5dkPkgIiKiYkTrcSCIiIhKKpZAqMcAgoiISII9wdTjL0RERESysQSCiIhIglUY6jGAICIikmAAoR6rMIiIiEg2lkAQERFJsARCPQYQREREEgwg1GMAQUREJMFunOrxFyIiIiLZWAJBREQkwSoM9RhAEBERSTCAUI9VGERERCQbSyCIiIgkWAKhHgMIIiIiCfbCUI+/EBEREcnGEggiIiIJVmGoxwCCiIhIggGEeqzCICIiItlYAkFERCTBEgj1GEAQERFJMIBQjwEEERGRBLtxqsdfiIiIiGRjCQQREZEEqzDUYwBBREQkwQBCPVZhEBERkWwsgSAiIpJgCYR6DCCIiIgk2AtDPf5CREREJBtLIIiIiCRYhaEeSyCIiIhINgYQREREJBurMIiIiCRYhaEeAwgiIiIJBhDqMYAgIiKSYAChHttAEBER6Ql/f380atQI1tbWcHBwQI8ePXDv3j3RNsnJyfDz84O9vT2srKzQu3dvREdHi7Z5+vQpunTpAgsLCzg4OOD777/H+/fvRducOXMGXl5eMDMzQ5UqVRAQECArrwwgiIiIJAwMDHT2kePs2bPw8/PDv//+ixMnTiAtLQ0dO3ZEYmKisM3YsWNx4MAB7N69G2fPnkVERAR69eolfJ+eno4uXbogNTUVFy5cwObNmxEQEICZM2cK24SFhaFLly5o06YNrl+/jjFjxmDo0KE4duyY5r+RUqlUyjq7AiKNjOjDxlHgKDsWJ5NUQV8T//zzj86O1bx5c633ffnyJRwcHHD27Fm0bNkScXFxKFu2LLZt24ZPP/0UAHD37l3UrFkTQUFBaNKkCY4cOYKuXbsiIiICjo6OAIA1a9Zg0qRJePnyJUxNTTFp0iQcOnQIN2/eFNLq27cvYmNjcfToUY3yxrs0ERGRnoqLiwMAlC5dGgAQHByMtLQ0tG/fXtimRo0acHNzQ1BQEAAgKCgIHh4eQvAAAD4+PoiPj8etW7eEbbIfI2ubrGNogo0oiYiIJHRZwpGSkoKUlBTROjMzM5iZmeW5X0ZGBsaMGYNmzZqhTp06AICoqCiYmprC1tZWtK2joyOioqKEbbIHD1nfZ32X1zbx8fFISkpCqVKl1J4XSyCIiIgkdNkGwt/fHzY2NqKPv7+/2jz4+fnh5s2b2LFjRyGcsXwsgSAiIipAU6ZMwbhx40Tr1JU+jBw5EgcPHsS5c+fg6uoqrHdyckJqaipiY2NFpRDR0dFwcnIStrl06ZLoeFm9NLJvI+25ER0dDYVCoVHpA8ASCCIiogJlZmYGhUIh+uQWQCiVSowcORL79u3DqVOnUKlSJdH3DRo0gImJCQIDA4V19+7dw9OnT+Ht7Q0A8Pb2RmhoKGJiYoRtTpw4AYVCgVq1agnbZD9G1jZZx9AEe2GQXmIvDMqOvTBIqqCvCTmNCdWR81AeMWIEtm3bhv/973+oXr26sN7GxkYoGRg+fDgOHz6MgIAAKBQKjBo1CgBw4cIFAJndOOvVqwcXFxcsWrQIUVFR8PX1xdChQ7FgwQIAmd0469SpAz8/PwwePBinTp3C6NGjcejQIfj4+GiUV70JINLS0oo6C6RHTExMijoLpEdSU1OLOgukZ0xNTQv0+EUVQOQWGG3atAlfffUVgMyBpMaPH4/t27cjJSUFPj4+WLVqlVA9AQDh4eEYPnw4zpw5A0tLSwwcOBALFy6EsfH/t1w4c+YMxo4di9u3b8PV1RUzZswQ0tAorwwgSB8xgKDsGECQVEEHEP/++6/OjtWkSROdHUufsBElERGRBKvN1GMAQUREJMEAQj22VCMiIiLZWAJBREQkwRII9RhAEBERSTCAUI9VGERERCQbSyCIiIgkWAKhHksgiIiISDYGEERERCQbqzCIiIgkWIWhHgMIIiIiCQYQ6rEKg4iIiGRjCQQREZEESyDUYwBBREQkwQBCPQYQREREEgwg1GMbCCIiIpKNJRBEREQSLIFQjwEEERGRBAMI9ViFQURERLKxBIKIiEiCJRDqsQSCiIiIZGMAQURERLJpFUC4u7vjv//+U1kfGxsLd3f3fGeKiIioKBkYGOjsU1Jp1QbiyZMnSE9PV1mfkpKCFy9e5DtTRERERakkP/h1RVYA8ddffwn/f+zYMdjY2AjL6enpCAwMRMWKFXWWOSIiItJPsgKIHj16AMiMzAYOHCj6zsTEBBUrVsSSJUt0ljkiIqKiwBII9WQFEBkZGQCASpUq4fLlyyhTpkyBZIqIiKgoMYBQT6s2EGFhYbrOBxERkd5gAKGe1gNJBQYGIjAwEDExMULJRJaNGzfmO2NERESkv7QKIObMmYO5c+eiYcOGcHZ2ZqRGREQlCp9r6mkVQKxZswYBAQHw9fXVdX6IiIioGNBqIKnU1FQ0bdpU13khIiKiYkKrAGLo0KHYtm2brvNCRESkFzgSpXpaVWEkJydj3bp1OHnyJDw9PWFiYiL6funSpTrJHBERUVEoyQ9+XdEqgAgJCUG9evUAADdv3hR9xx+diIio5NMqgDh9+rSu80FERETFSL6m83748CGOHTuGpKQkAIBSqdRJpoiIiIoS20Cop1UA8d9//6Fdu3aoVq0aOnfujMjISADAkCFDMH78eJ1mkIiIiPSPVgHE2LFjYWJigqdPn8LCwkJY//nnn+Po0aM6yxwREVFRYAmEelq1gTh+/DiOHTsGV1dX0fqqVasiPDxcJxkrCmlpaXj8+DEiIiIQHR2Nd+/eIS0tDVZWVrC1tUW1atXg7u4OIyOjos6q1iIiInDr1i1ERkYiKSkJJiYmKF26NNzc3FCjRg1RQEi68/jxY9y9exfR0dFISkqCubk57O3tUalSJdSoUQOmpqZFnUXSQPZ7RExMDBITE/H+/XtYWlqWmHsEZSrJD35d0SqASExMzPFB8/r1a5iZmeU7U4Xp+PHjCAoKwvXr1xEWFob379/nub21tTU+/vhjfPnll3B3d5edXnp6Oh48eIDQ0FDh8+jRI9F8IsOHD4efn5/sY+cmLS0Nu3fvxu7du/HgwYNctzMyMkKNGjXQsmVLDB8+HIaG+Woi88FLSEjA77//jt27d+P58+e5bmdiYgJPT0/4+Phg4MCB+U733r176N27N9LS0kTr/f390atXr3wf/0Nz/Phx/Pvvv7h27RqePHmi8T2if//+su4RPj4+iIiI0DqfGzduRKNGjdRut3//fsyYMUPrdBo2bIhNmzZpvT+VHFoFEC1atMCWLVswb948AJmRWkZGBhYtWoQ2bdroNIMF7ccff0R0dLTG2799+xa7du3C3r17MWzYMIwYMUKjSHXjxo04c+YM7ty5IzQ6LQw3btzA5MmT8ezZM7Xbpqen49atW7h16xaGDh1a7IJBfXL69GlMnz4dr169UrttWloagoODER4enu8AIj09HdOmTVMJHkh7P/74I2JiYjTeXnqPGD58ON9mqUTSKoBYtGgR2rVrhytXriA1NRUTJ07ErVu38Pr1a5w/f17XeSx0ZmZmcHZ2hpWVFTIyMhAbG4vIyEhRL5P3799j9erViIqKEgKpvBw6dAj37t0ryGyrOHr0KCZPnqzyxmRsbAxHR0fY2dkhLS0N//33n0YPOtJMQEAAFi5cqNIryczMDA4ODrCzs0NycjJevnyJN2/e6Dzt0NBQnR6TVMm9R8ydO7cIc0vaYNCnnlYBRJ06dXD//n2sWLEC1tbWSEhIQK9eveDn5wdnZ2dd57HAOTg4oGXLlmjQoAHq1q0LV1dXleL7uLg4nDhxAqtXrxaVWOzbtw9eXl7o2bOnVmkbGhrC1NQUycnJ+ToHqfPnz6sED3Xr1sVXX32FZs2aqVRBxcbG4vLlyzh9+jROnDih07x8SHbv3g1/f3/RupYtW8LX1xdNmjRRaesQHR2Nf//9F4GBgQgJCclX2k+fPsXy5cuFZQsLC7x79y5fx6RMDg4OaNGiBRo0aIB69eqhXLlyud4j1qxZo3KPqF+/vqx7hL29vcp1pE61atVkbZ9l0KBB8Pb21nh7hUKhVTrFDQMI9QyUejJ4Q1EVud67dw/VqlXT+GKJi4vDsGHDcPv2bWFd2bJlERgYmGebgd69e+PevXtwcXFBnTp1hE/t2rWxYMEC/O9//xO2zW8biLi4OHTr1g2vX78W1k2YMAEDBw7U6Dzfvn0LKyurIv0Dkg6PXhyEh4ejW7duSElJAZB5DgsXLkTXrl012j8uLg42NjZapa1UKjFw4EBcvHgRANCmTRskJibi0qVLwjbFuQ1EampqkaWt7T3izp07wrqyZcvi5MmTed4jsreBcHFxwbFjx/KX8VxI20DMmzcPPXr0KJC0ClJBNzx++vSpzo7l5uams2PpE61KIIDM+TBCQkIQExMjagAIAJ988km+M1ZYqlevLmt7Gxsb+Pv7o0ePHkJx5cuXL3Ht2jU0aNAg1/1mzJiB8uXLw97ePl/51cTixYtFwcO4cePw1Vdfaby/tbV1AeSq5JsxY4YQPACZ/w4ff/yxxvtrGzwAwK5du4TgwcLCArNmzcLEiRO1Ph79P23uEQsXLpR9jyAqbrQKII4ePYoBAwbkWG9uYGCA9PT0fGdMn1WuXBm1atXCrVu3hHWPHz/O8+aQNXdIQbt79y72798vLHt6esoKHkg7J0+eFB7gANCpUydZwUN+REdHY/HixcLy2LFji2VVYkni7u4u+x5B+oVVGOpp1U9v1KhR6NOnDyIjI5GRkSH6lPTgIUv58uVFy7puDKetP//8U7T8zTffsDtmIdi1a5doeeTIkYWW9uzZs/H27VsAmQHjl19+WWhpU+6k94jY2NiiyQhRAdGqBCI6Ohrjxo2Do6OjrvNTbEjrZPWhYVFycjIOHz4sLNvZ2aFZs2ZFmKMPQ3R0NP755x9huWbNmqhatWqhpH3o0CGcOnUKQGbvmnnz5jFg1BPZq7MAVg0WNyyBUE+rO82nn36KM2fO6DgrxYdSqVSZxrxWrVpFlJv/d/36dcTHxwvLXl5eMDbWupkLaejcuXOikrfGjRsXSrpv3rzBDz/8ICwPGjQINWrUKJS0KW9KpVJUfQHoxz2CNMehrNXT6umyYsUK9OnTB3///Tc8PDxUWsyPHj1aJ5nTV3v37hUNLFOpUiV4eHgUYY4ySYOa7I2/wsPDsW/fPly4cAGRkZFITEyEnZ0dXF1d4e3tjU6dOqFixYqFnOOSQTruQvaH+O3bt7F3715cunQJkZGRSE1Nhb29PSpUqIDmzZuja9euWpfkzZ8/X2gs6+bmVqjVJpS3ffv26eQe8ebNG0RHRyMhIUEYUt/R0VHnD6X09HQ8f/5cqIq1tbVF2bJlYWlpqdN0qGTRKoDYvn07jh8/DnNzc5w5c0Z0MRsYGJToAOJ///uf6K3P0NAQ06ZN04soUxpAuLq6Ii0tDStXrsTGjRtVestER0cjOjoawcHBWL16NXr16oUJEybwpiGT9HcvX748EhMTMX/+fOzZs0dl+xcvXuDFixe4cOECli9fjoEDB2LUqFGyuq6ePXsWBw4cEJbnzJkDc3Nz7U+CdCane8TUqVNl3SNev36N7t274/Hjxyrf2djYwMvLC507d0aHDh3yPe/G+vXrsXDhQiQmJorWGxoaomrVqvD29ka/fv3g4uKSr3So5NEqgJg2bRrmzJmDyZMnl7j61idPngjTkwOZo8nFx8fjwYMHOH36NB49eiR8Z2JigtmzZ6NJkyZFkVUV0n7LpUqVwqhRo0T187lJT0/H7t27ERoaitWrV6Ns2bIFlc0SRzqBnIGBAb788kvRWCG5SU5Oxtq1axEaGopff/0VVlZWavdJSEjArFmzhOXu3bujadOm8jNOWsntHvHw4cMc7xGzZs2SfY9ITk7OMXgAMseZOH36NE6fPo3y5ctjzpw5Gs2BkZvcJkDMyMjAvXv3cO/ePfz+++/o1asXvv/++w8mUNWHl0J9p1UAkZqais8//7zEBQ8AsGPHDvz+++95bmNgYIDmzZvju+++06s656yW+Fl+++030UOsYcOGaN++PVxdXZGamoqHDx/i0KFDohvI3bt38d1332HLli1sP6GBjIwMlTe3+fPnC7+7gYEBWrdujVatWsHJyQlJSUm4ffs2/ve//4mKuC9cuIApU6bg119/VZvmTz/9JDzA7OzsMHnyZB2eEamzc+dOje4RzZo1w5gxY2SPIyHHs2fPMGzYMEyYMKFAe9+8f/8eu3btwrVr17B27doP4gWDAYR6Wj0hBg4ciJ07d2Lq1Km6zk+x0LFjR/Tv31+vggdANYDIeoiZmZlh4cKF6NChg+j7jh07YtiwYfjll1+wefNmYX1ISAh+++03DB8+vOAzXcy9fftWZc6LrMZztra2WLlyJRo2bCj6vnPnzhg+fDhmzpyJgwcPCuuPHz+O/fv35zkq4JUrV7Bjxw5hefLkyShdurQOzoR0KeseITd4sLKyQrNmzdCsWTNUq1YNrq6usLS0RHJyMmJiYnDjxg3s378fV69eFfZJT0/HokWLYG9vL2vskayZd+vVq4fKlSvD1tYWJiYmiIuLw4MHD3DhwgXs2bNHdF958OABRo4ciU2bNuU4IzN9WLQqQsi6YFu1aoVRo0Zh3Lhxok9Jd+zYMQwYMAADBw7U6XCn+ZXbvAf+/v4qwUMWExMTfP/99ypDHG/dupXzKGggt9/IyMgI69atUwkeslhaWmLx4sVo3ry5aP2aNWtUApIsKSkpmDZtmvB906ZNi+UQxB8Cbe4R48aNQ2BgIH766Sf07NkTtWvXho2NDYyNjWFlZQV3d3f07NkTmzdvxi+//CLqFqpUKjFr1iyNJsWrWbMmdu7cid27d2PUqFFo0aIFXFxcYGFhARMTE5QpUwbe3t4YP348jhw5grZt24r2v337NlauXCnvBymGiqoXxrlz59CtWze4uLjAwMBANDAgkPlvPXPmTDg7O6NUqVJo3749Hjx4INrm9evX6N+/PxQKBWxtbTFkyBAkJCSItgkJCUGLFi1gbm6O8uXLY9GiRbJ/I60CiNDQUNSvXx+Ghoa4efMmrl27JnyuX7+uzSH1xuTJk3Hz5k3hc+XKFZw8eRIrV65Er169RPV/wcHB6Nu3r0ojuqKS0/TbzZo1Q8eOHdXuO3HiRFH9e3x8PI4fP67T/JVEuU15/umnn6Ju3bp57mtoaIjZs2eLqgLDwsJE81dkt3z5cjx58gQAYG5ujjlz5miXacqXSZMmITQ0VPhcvnwZJ06cyPEecfXqVfTt21elS2dOfHx8NH6rb9euHVavXi1KKykpCevWrVO7b/Xq1TXuUmpjY4Off/4Z7du3F63fuXOnaMIw0p3ExETUrVs31yBt0aJFWL58OdasWYOLFy/C0tISPj4+ogkZ+/fvj1u3buHEiRM4ePAgzp07h6+//lr4Pj4+Hh07dkSFChUQHByMxYsXY/bs2RpdP9lpFUBkNeDJ6ZM1qE1JYW5uDicnJ7Rq1Qpz587F3r17RVUX8fHx+O6770TjLxSVUqVKqaz79NNPNdrXyspKpfjzypUrOslXSZbbDf+zzz7TaP/y5curNIC8fPmyyna3bt3Cpk2bhGU/P78SO0FPcZN1j2jZsiXmzJmDPXv2iO4Rb9++xejRo3V+j6hbty4GDRokWnf48GGV3lb5lRXoZi/xSElJ4QtGAfn444/xww8/5Dh7q1KpxC+//ILp06eje/fu8PT0xJYtWxARESGUVNy5cwdHjx7F+vXr0bhxYzRv3hy//vorduzYIUzW9scffyA1NRUbN25E7dq10bdvX4wePRpLly6VlVedtIKMj4/H/v37cffuXV0cTq+5ubnht99+g5OTk7AuOjpadHMvKjk9zHIrQs+JdFtN3po+dObm5ird6CwtLWUNGiRtQS8t0Xr//j2mTp0qDFZVvXp1DB48WMscU0Fzc3PDunXrRPeImJgYBAQE6Dyt/v37i66/uLi4Avm7tbGxUXmgXbhwQefp6BN9HEgqLCwMUVFRohIhGxsbNG7cGEFBQQCAoKAg2Nraiu7n7du3h6GhoTBfT1BQEFq2bCma0dTHxwf37t2TNS2DVgHEZ599hhUrVgDILDZr2LAhPvvsM3h4eOTY710qJSUF8fHxoo902Fd9ZmdnpzLddvbpuIuKdKZPhUIBOzs7jfeXDiT133//6SJbJZ60EWOFChVk9VCqVKmSaFn6u//1119CcG5oaIh58+axh4yes7Ozw4gRI0TrCuIeYWNjg5o1a4rWZVVz6Zp0hNWCSkdf6DKA0NUzLyoqCgBUBp9zdHQUvouKioKDg4Poe2NjY5QuXVq0TU7HyJ6GJrQKIM6dO4cWLVoAyBxxTalUIjY2FsuXLxcNoJIbf39/2NjYiD4//vijNlkpMu3atRNFljExMULxUFFxd3cXLWsypkBe2+tDtUxxULlyZdGyrn/37MsZGRn47LPPUL16dbUfaVuKKVOmiL6XPuBItwrrHpG9pAOAMDppQaejLxMIFgc5PfP8/f2LOlv5plUAERcXJ7x1HT16FL1794aFhQW6dOmi0ho0J1OmTEFcXJzoM2nSJG2yUmQUCgVsbGxE6zRpAV2QpA8y6YRf6kgj4pzaVJCqKlWqiJbz+7t/KAP1lHSFdY+Qlka9f/9e52kUZjolUU7PvClTpsg+TlYQJ23AGh0dLXzn5OQkGmMGyPy3ev36tWibnI6RPQ1NaBVAlC9fHkFBQUhMTMTRo0eFVv5v3rzR6OZnZmYGhUIh+uTWmr04KepiZWm9e1xcnKw/cumbi/TmRzmT/u5yHxLSKgtbW9v8Zon0VEHcI6TXW0GNCyK9TuVUjxZHuqzC0NUzr1KlSnByckJgYKCwLj4+HhcvXoS3tzcAwNvbG7GxsQgODha2OXXqFDIyMoRqKG9vb5w7dw5paWnCNidOnED16tVl/btqdTWPGTMG/fv3h5WVFSpUqIDWrVsDyKza0IdJpQpDYmIi4uLiROukbRAKm5eXFxQKhVDknZaWhgcPHqjUkeZGOvSytG6ecta6dWsYGhoKrd+fP3+O2NhYjQMBaaM36e/u4+ODatWqyc7XwoULce/ePWF5yJAhonEnOABVwSqMe0RqaqrK9aPt5GzqZB+8CpD3pkqaS0hIwMOHD4XlsLAwXL9+HaVLl4abmxvGjBmDH374AVWrVkWlSpUwY8YMuLi4CGPC1KxZE506dcKwYcOwZs0apKWlYeTIkejbt68wn8kXX3yBOXPmYMiQIZg0aRJu3ryJZcuW4eeff5aVV60CiBEjRuCjjz7Cs2fP0KFDB6HBmLu7u0ZtIEqCs2fPigb8KV26dJEP72psbIxWrVqJJlk6ceKExgHEsWPHRMv5GV//Q2Jvbw8vLy9Rt9cTJ06gT58+avd9//49Tp48KVr30UcfiZadnZ3h7OwsO1/SEqQqVapwzoxCdO7cuQK/Rxw5cgRJSUnCsqmpKerXr6/TNIDM6/Svv/4SrSusaeuLSlENZX3lyhW0adNGWM4anHHgwIEICAjAxIkTkZiYiK+//hqxsbFo3rw5jh49Kir9/+OPPzBy5Ei0a9cOhoaG6N27N5YvXy58b2Njg+PHj8PPzw8NGjRAmTJlMHPmTNFYEZrQujytYcOGQjeR9PR0hIaGomnTpiW+WAvInOhm1apVonUtW7bUi7lB+vTpIwogdu3aBV9fX7X/Ln///bfoTcbIyEhl8BjK3eeffy4KINavX4/u3buLuknlZNeuXXj58qWwbGVlJTRQpuKrMO4Rr169Ej0UgMzRSQui7dKGDRtURtTM/pAriYoqgGjdunWuo9ECmfmaO3cu5s6dm+s2pUuXxrZt2/JMx9PTE3///bfW+QS0bAMxZswYbNiwAUBm8NCqVSt4eXmhfPnyOHPmTL4yVJiWLFmC0NBQWfvExcVh5MiRoi5MRkZGGDBggI5zpx0vLy+0bNlSWI6NjcXEiRPz7DL09OlTzJw5U7SuY8eOHKhIhq5du4qqGZ48eYKZM2fmOajPjRs3sHjxYtG6L774QjRgDxWtpUuXyh5pNi4uDqNGjVK5R/j6+ua4/cuXL7Fy5UqV6o68vHjxAt9++62osZyBgYHa+Wt27twpjAWgqc2bN6uMiti6dWvUrl1b1nGo5NGqBOLPP/8UZn47cOAAwsLCcPfuXWzduhXTpk3D+fPndZrJgnLhwgVs2rQJHh4e6NSpExo3bozKlSvDxMREtJ1SqURYWBiOHz+OP/74Q6X7kq+vr9o66ri4uFynd87+Bgpk1qFnDQoiVatWLbWNG8ePH49Lly4JQ5sGBQXB19cXEyZMQKNGjYTIOjk5GUePHsXSpUtFDShLly6NCRMm5JkGiRkaGmLq1KkYNGiQ8Pawb98+REdHY/z48ahTp46w7du3b/Hnn39i+fLlork0KlasiG+++abQ8065y36P8PHx0egesW3bNln3iNTUVKxZswZbtmxB69at0aFDB3h6eqr05Qcyg/39+/djx44dKpPn9e/fX+0AZiEhIfjhhx9QvXp1+Pj4oFmzZqhSpYpKSVlSUhIuXryIgIAAUWM8ILOR78SJE/NMhz4MBsq8ykpyYW5ujocPH8LV1RVff/01LCws8MsvvyAsLAx169bVavyA7K1BC0vv3r1FjcyAzMmlHB0dYW1tDRMTEyQmJiIqKkplyuYs3bt3x7x589QWTV66dEknowdu3LhRpY48J4cPH8bkyZNV3oBtbW3h4uKCtLQ0PH/+XFR/CmT+265YsQJNmjTJd17zQ3qDLi7WrVuHJUuWqKwvW7YsHB0dkZSUhKdPn6pc77a2ttiyZYtOp3729fUVjQXh7++vMmlacSG3a6yufPrppzneIxwcHKBQKDS6R3zyySd53iNevHiBTp06qay3tbVF6dKlYWVlheTkZLx69SrXMR46duyIxYsXq70PTZs2TaUtg7GxseieFx8fjxcvXuTYg8vS0hJr1qxBvXr18kynMKirHswvXY6nUVIbLGtVAuHo6Ijbt2/D2dkZR48exerVqwFkzkwoHda3uMl6sKpjZWWFMWPG4PPPP9fLeeM7d+4MY2NjTJ8+XfSWGxsbi9jY2Bz3KVu2LJYtWwZPT89CymXJ8/XXX6NUqVL48ccfRUHCy5cvVUqaslSqVAlr1qxRGQmU9FNaWhpevHiBFy9e5Lld1j3is88+0+oekdffahZTU1OMHj0aAwYM0Po+9P79e7XnAmTWmS9cuBDly5fXKp3iRh/v6/pGqwBi0KBB+Oyzz+Ds7AwDAwOhsd3FixdFk8jou0WLFuHs2bO4cOECbt68qTLdqZSBgQGqVq2Kbt26oXv37nofVXbs2BGenp5Ys2YNDh06pFLakKVs2bLo06cPvvrqK41nA6Tc+fr6omnTplixYgVOnDiRa+maq6srBg4ciL59+xb42xRpZ9GiRThz5gyCgoJk3yM++eQTje4R9vb2mDx5Mi5duoQbN25oNIS8i4sLunbtis8//zzHqo7cfP7551AoFAgODsaDBw/UjhNjZmaGhg0bol+/fmjZsiUfqiSiVRUGkNkO4tmzZ+jTpw9cXV0BZDa2sbW1Rffu3WUfryiqMLLLyMhAeHg4nj59isjISCQmJuL9+/ewsLCAtbU1XFxcUKtWLdnDFOuL5ORk3LhxA2FhYYiLi4O5uTns7OxQrVo1vQz6imsVhlRCQgKuXr2K8PBwvH37FhYWFihTpgxq1aqlMvQ45a6oqjCyy7pHPHv2DJGRkUhISMD79+9haWkJKysrlCtXDjVr1sz3PeLly5d48uQJIiMj8ebNGyQnJ8PExAQKhQKlS5dGnTp1ZAUNuUlLS8Pjx4/x/PlzvHz5UrjnWVlZQaFQoEKFCqhZs6be/i0WdNCty6G6S2rvRK0DCF0r6gCC9Iu+3rSoaOhDAEH6paADCHXVR3KU1NFltR4HIjExEWfPnsXTp09V/rhHjx6d74wRERGR/tKqBOLatWvo3Lkz3r17h8TERJQuXRqvXr2ChYUFHBwc8PjxY9kZYQkEZccSCMqOJRAkVdAlEHLG5VCnpM4rpNVAUmPHjkW3bt3w5s0blCpVCv/++y/Cw8PRoEED/PTTT7rOIxEREekZrQKI69evY/z48TA0NISRkRFSUlJQvnx5LFq0CFOnTtV1HomIiEjPaBVAmJiYCAOWODg4CGOk29jY4NmzZ7rLHRERURHQ5XTeJZVWjSjr16+Py5cvo2rVqmjVqhVmzpyJV69eYevWraIhe4mIiKhk0qoEYsGCBcL0wvPnz4ednR2GDx+Oly9fYt26dTrNIBERUWFjCYR6snthKJVKPHv2DA4ODqL5x/OLvTAoO/bCoOzYC4OkCroXhnSysvwoqTPsyi6BUCqVqFKlCts6EBERfcBkBxCGhoaoWrWqRuO1ExERFUeswlBPqzYQCxcuxPfff4+bN2/qOj9ERERUDGg1EqWdnR3evXuH9+/fw9TUFKVKlRJ9r8086mwDQdmxDQRlxzYQJFXQbSDUzbwqR3GdhFEdrbpx/vLLLzrOBhERkf4oyVUPusLZOEkvsQSCsmMJBEkVdAlEYmKizo5laWmps2PpE61n48ySnJys8setUCjye1giIiLSY1o1okxMTMTIkSPh4OAAS0tL2NnZiT5ERETFGXthqKdVADFx4kScOnUKq1evhpmZGdavX485c+bAxcUFW7Zs0XUeiYiISM9o1QbCzc0NW7ZsQevWraFQKHD16lVUqVIFW7duxfbt23H48GHZGWEbCMqObSAoO7aBIKmCbgPx7t07nR3LwsJCZ8fSJ1qVQLx+/Rru7u4AMts7ZHXbbN68Oc6dO6e73BERERUBVmGop1UA4e7ujrCwMABAjRo1sGvXLgDAgQMHYGtrq7PMERERkX7SKoAYNGgQbty4AQCYPHkyVq5cCXNzc4wdOxbff/+9TjNIRERE+kcn40CEh4cjODgYVapUgaenp1bHYBsIyo5tICg7toEgqYJuA5GcnKyzY+ly5mp9ovU4EIGBgQgMDERMTAwyMjJE323cuDHfGSMiIiL9pVUAMWfOHMydOxcNGzaEs7NziW4kQkRERKq0qsJwdnbGokWL4Ovrq7OMsAqDsmMVBmXHKgySYhVG0dOqBCI1NRVNmzbVdV6IiIj0AkvW1dOqF8bQoUOxbds2XeeFiIiIigmNSyDGjRsn/H9GRgbWrVuHkydPwtPTU6W4eenSpbrLIREREekdjQOIa9euiZbr1asHALh586ZoPYt9iIiouOOzTD2djAOhC2xESdmxESVlx0aUJFXQjSh1ec0VdF6LilZtIIiIiOjDpvVAUkRERCUVqzDUYwkEERERycYAgoiIiGRjFQYREZEEqzDUYwkEERERycYAgoiIiGRjFQYREZEEqzDUYwkEERERycYAgoiIiGRjFQYREZEEqzDUYwkEERERycYAgoiIiGRjFQYREZEEqzDUYwkEERERycYSCCIiIgmWQKjHEggiIiKSjQEEERERyWagVCqVRZ0JypSSkgJ/f39MmTIFZmZmRZ0dKmK8Hig7Xg+kbxhA6JH4+HjY2NggLi4OCoWiqLNDRYzXA2XH64H0DaswiIiISDYGEERERCQbAwgiIiKSjQGEHjEzM8OsWbPYQIoA8HogMV4PpG/YiJKIiIhkYwkEERERycYAgoiIiGRjAEFERESyMYD4AFSsWBG//PJLUWfjg9a6dWuMGTOmqLNBGuC/FZFmGEAQFQMBAQGwtbUt6mwQqTAwMMD+/fuLOhtUBBhA6IHU1NSizgIRFZD09HRkZGQUdTaIdI4BhBZat26N0aNHY+LEiShdujScnJwwe/Zs4funT5+ie/fusLKygkKhwGeffYbo6Gjh+9mzZ6NevXpYv349KlWqBHNzcwCZkfzatWvRtWtXWFhYoGbNmggKCsLDhw/RunVrWFpaomnTpnj06JFwrEePHqF79+5wdHSElZUVGjVqhJMnTxbab0Gay8jIyPWaWbp0KTw8PGBpaYny5ctjxIgRSEhIAACcOXMGgwYNQlxcHAwMDGBgYCDsm5KSggkTJqBcuXKwtLRE48aNcebMmcI/uRJG238r4P9Li/766y/UqlULZmZmePr0KS5fvowOHTqgTJkysLGxQatWrXD16lVhP6VSidmzZ8PNzQ1mZmZwcXHB6NGjhe9XrVqFqlWrwtzcHI6Ojvj000+F71q3bo1Ro0ZhzJgxsLOzg6OjI3777TckJiZi0KBBsLa2RpUqVXDkyBHRed68eRMff/wxrKys4OjoCF9fX7x69Up03LzudRUrVgQA9OzZEwYGBsIyfSCUJFurVq2UCoVCOXv2bOX9+/eVmzdvVhoYGCiPHz+uTE9PV9arV0/ZvHlz5ZUrV5T//vuvskGDBspWrVoJ+8+aNUtpaWmp7NSpk/Lq1avKGzduKJVKpRKAsly5csqdO3cq7927p+zRo4eyYsWKyrZt2yqPHj2qvH37trJJkybKTp06Cce6fv26cs2aNcrQ0FDl/fv3ldOnT1eam5srw8PDhW0qVKig/Pnnnwvr56Ec5HXNKJVK5c8//6w8deqUMiwsTBkYGKisXr26cvjw4UqlUqlMSUlR/vLLL0qFQqGMjIxURkZGKt++fatUKpXKoUOHKps2bao8d+6c8uHDh8rFixcrzczMlPfv3y+ycy3u8vNvpVQqlZs2bVKamJgomzZtqjx//rzy7t27ysTERGVgYKBy69atyjt37ihv376tHDJkiNLR0VEZHx+vVCqVyt27dysVCoXy8OHDyvDwcOXFixeV69atUyqVSuXly5eVRkZGym3btimfPHmivHr1qnLZsmWiPFtbWyvnzZunvH//vnLevHlKIyMj5ccff6xct26d8v79+8rhw4cr7e3tlYmJiUqlUql88+aNsmzZssopU6Yo79y5o7x69aqyQ4cOyjZt2mj8W8TExCgBKDdt2qSMjIxUxsTEFOw/DukVBhBaaNWqlbJ58+aidY0aNVJOmjRJefz4caWRkZHy6dOnwne3bt1SAlBeunRJqVRmBhAmJiYqf2wAlNOnTxeWg4KClACUGzZsENZt375daW5unmf+ateurfz111+FZQYQRS+vayYnu3fvVtrb2wvLmzZtUtrY2Ii2CQ8PVxoZGSlfvHghWt+uXTvllClTdJPxD5Au/q0AKK9fv55nOunp6Upra2vlgQMHlEqlUrlkyRJltWrVlKmpqSrb7tmzR6lQKIRgQ12e379/r7S0tFT6+voK6yIjI5UAlEFBQUqlUqmcN2+esmPHjqLjPHv2TAlAee/evRyPq1Sq/hYAlPv27cvzXKlkYhWGljw9PUXLzs7OiImJwZ07d1C+fHmUL19e+K5WrVqwtbXFnTt3hHUVKlRA2bJl8zyuo6MjAMDDw0O0Ljk5GfHx8QCAhIQETJgwATVr1oStrS2srKxw584dPH36VDcnSjqT2zUDACdPnkS7du1Qrlw5WFtbw9fXF//99x/evXuX6/FCQ0ORnp6OatWqwcrKSvicPXtWVM1F8uX338rU1FTlGNHR0Rg2bBiqVq0KGxsbKBQKJCQkCH+rffr0QVJSEtzd3TFs2DDs27cP79+/BwB06NABFSpUgLu7O3x9ffHHH3+oXBvZ0zMyMoK9vb3KvQOAcB43btzA6dOnRddOjRo1AEB0/eT1W9CHjQGElkxMTETLBgYGshpKWVpaqj2ugYFBruuy0powYQL27duHBQsW4O+//8b169fh4eHBhpl6KLdr5smTJ+jatSs8PT2xZ88eBAcHY+XKlQDybmCbkJAAIyMjBAcH4/r168Lnzp07WLZsWYGeS0mX33+rUqVKCX+rWQYOHIjr169j2bJluHDhAq5fvw57e3thv/Lly+PevXtYtWoVSpUqhREjRqBly5ZIS0uDtbU1rl69iu3bt8PZ2RkzZ85E3bp1ERsbm2ee87p3JCQkoFu3bqJr5/r163jw4AFatmyp9rcgMi7qDJQ0NWvWxLNnz/Ds2TOhFOL27duIjY1FrVq1dJ7e+fPn8dVXX6Fnz54AMm8KT5480Xk6VHCCg4ORkZGBJUuWwNAwM6bftWuXaBtTU1Okp6eL1tWvXx/p6emIiYlBixYtCi2/HzJN/q1yc/78eaxatQqdO3cGADx79kzUYBHIDDy6deuGbt26wc/PDzVq1EBoaCi8vLxgbGyM9u3bo3379pg1axZsbW1x6tQp9OrVS6tz8fLywp49e1CxYkUYG2v/KDAxMVG5NunDwBIIHWvfvj08PDzQv39/XL16FZcuXcKAAQPQqlUrNGzYUOfpVa1aFXv37sX169dx48YNfPHFF3w7KGaqVKmCtLQ0/Prrr3j8+DG2bt2KNWvWiLapWLEiEhISEBgYiFevXuHdu3eoVq0a+vfvjwEDBmDv3r0ICwvDpUuX4O/vj0OHDhXR2ZRsmvxb5aZq1arYunUr7ty5g4sXL6J///4oVaqU8H1AQAA2bNiAmzdv4vHjx/j9999RqlQpVKhQAQcPHsTy5ctx/fp1hIeHY8uWLcjIyED16tW1Phc/Pz+8fv0a/fr1w+XLl/Ho0SMcO3YMgwYNkhUQVKxYEYGBgYiKisKbN2+0zg8VPwwgdMzAwAD/+9//YGdnh5YtW6J9+/Zwd3fHzp07CyS9pUuXws7ODk2bNkW3bt3g4+MDLy+vAkmLCkbdunWxdOlS/Pjjj6hTpw7++OMP+Pv7i7Zp2rQpvv32W3z++ecoW7YsFi1aBADYtGkTBgwYgPHjx6N69ero0aMHLl++DDc3t6I4lRJPk3+r3GzYsAFv3ryBl5cXfH19MXr0aDg4OAjf29ra4rfffkOzZs3g6emJkydP4sCBA7C3t4etrS327t2Ltm3bombNmlizZg22b9+O2rVra30uLi4uOH/+PNLT09GxY0d4eHhgzJgxsLW1FUpXNLFkyRKcOHEC5cuXR/369bXODxU/nM6biIiIZGMJBBEREcnGAIKIiIhkYwBBREREsjGAICIiItkYQBAREZFsDCCIiIhINgYQREREJBsDCCIiIpKNAQQRERHJxgCCiIiIZGMAQURERLIxgCAiIiLZ/g/DkXVqOlDWZwAAAABJRU5ErkJggg==\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "test_X_2k7, test_y_2k7 = pre_process_features(test['text'], test['label'], tokenized=True, lowercased = True)\n",
        "test_X_2k7, test_y_2k7 = make_featues(test_X_2k7, test_y_2k7, tokenizer, is_one_hot_label=False)"
      ],
      "metadata": {
        "id": "UL5ObppdMhDJ"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "prediction_2k7 = model.predict(test_X_2k7, batch_size=batch_size, verbose=0)\n",
        "y_pred_2k7 = prediction_2k7.argmax(axis=-1)\n",
        "\n",
        "cf2k7 = confusion_matrix(test_y_2k7, y_pred_2k7)"
      ],
      "metadata": {
        "id": "lcDY6TyfPYyo"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "accuracy_score(test_y_2k7, y_pred_2k7)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "t2OOP1RYPdPw",
        "outputId": "7266ffb8-65ab-4c3c-b0a4-c8ee6ac6185e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.8277777777777777"
            ]
          },
          "metadata": {},
          "execution_count": 41
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "f1_score(test_y_2k7, y_pred_2k7, average='macro')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "A18dctrfPsJz",
        "outputId": "8ab0574d-0160-4e40-9d57-ccae160de735"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.6058923714380131"
            ]
          },
          "metadata": {},
          "execution_count": 42
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "df_cm2k7 = pd.DataFrame(cf2k7, index = [\"normal\",\"hate\",\"harassment\"],\n",
        "                  columns = [\"normal\",\"hate\",\"harassment\"])\n",
        "plt.clf()\n",
        "sn.heatmap(df_cm2k7, annot=True, cmap=\"Greys\",fmt='g', cbar=True, annot_kws={\"size\": 30})"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 447
        },
        "id": "8N1U2BU1Pvc0",
        "outputId": "afc5e5bd-26d1-48e8-8e91-56521c6884dc"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<Axes: >"
            ]
          },
          "metadata": {},
          "execution_count": 43
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## GRU"
      ],
      "metadata": {
        "id": "qUNZtII7tv91"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "from keras.layers import Dense, Input, Bidirectional, GRU\n",
        "from keras.layers import GlobalMaxPooling1D, GlobalAveragePooling1D, concatenate, SpatialDropout1D"
      ],
      "metadata": {
        "id": "UgDzBDtiLina"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "input = Input(shape=(sequence_length,))\n",
        "x = Embedding(num_words, embedding_dim, weights=[embedding_matrix])(input)\n",
        "x = SpatialDropout1D(0.2)(x)\n",
        "x = Bidirectional(GRU(80, return_sequences=True))(x)\n",
        "avg_pool = GlobalAveragePooling1D()(x)\n",
        "max_pool = GlobalMaxPooling1D()(x)\n",
        "conc = concatenate([avg_pool, max_pool])\n",
        "output = Dense(3, activation=\"softmax\")(conc)\n",
        "\n",
        "# this creates a model that includes\n",
        "model = Model(inputs=input, outputs=output)\n",
        "\n",
        "model.summary()"
      ],
      "metadata": {
        "id": "wSU87vShtzmF",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "acdd15c2-0e7a-4503-b16c-3ee241aba3b9"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model_1\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                Output Shape                 Param #   Connected to                  \n",
            "==================================================================================================\n",
            " input_2 (InputLayer)        [(None, 100)]                0         []                            \n",
            "                                                                                                  \n",
            " embedding_1 (Embedding)     (None, 100, 300)             6449700   ['input_2[0][0]']             \n",
            "                                                                                                  \n",
            " spatial_dropout1d (Spatial  (None, 100, 300)             0         ['embedding_1[0][0]']         \n",
            " Dropout1D)                                                                                       \n",
            "                                                                                                  \n",
            " bidirectional (Bidirection  (None, 100, 160)             183360    ['spatial_dropout1d[0][0]']   \n",
            " al)                                                                                              \n",
            "                                                                                                  \n",
            " global_average_pooling1d (  (None, 160)                  0         ['bidirectional[0][0]']       \n",
            " GlobalAveragePooling1D)                                                                          \n",
            "                                                                                                  \n",
            " global_max_pooling1d (Glob  (None, 160)                  0         ['bidirectional[0][0]']       \n",
            " alMaxPooling1D)                                                                                  \n",
            "                                                                                                  \n",
            " concatenate_1 (Concatenate  (None, 320)                  0         ['global_average_pooling1d[0][\n",
            " )                                                                  0]',                          \n",
            "                                                                     'global_max_pooling1d[0][0]']\n",
            "                                                                                                  \n",
            " dense_1 (Dense)             (None, 3)                    963       ['concatenate_1[0][0]']       \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 6634023 (25.31 MB)\n",
            "Trainable params: 6634023 (25.31 MB)\n",
            "Non-trainable params: 0 (0.00 Byte)\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "batch_size = 256\n",
        "epochs = 40\n",
        "\n",
        "MODEL_FILE = 'GRU_model_v1.h5'\n",
        "adam = Adam(learning_rate=1e-4, beta_1=0.9, beta_2=0.999, epsilon=1e-08)\n",
        "\n",
        "model.compile(optimizer=adam, loss='categorical_crossentropy', metrics=['accuracy'])\n",
        "\n",
        "model.fit(train_X, train_y, batch_size=batch_size, epochs=epochs, verbose=1, validation_data=(dev_X, dev_y))\n",
        "model.save(DATA_PATH + MODEL_FILE)\n",
        "\n",
        "prediction = model.predict(test_X, batch_size=batch_size, verbose=0)\n",
        "y_pred = prediction.argmax(axis=-1)\n",
        "\n",
        "cf2 = confusion_matrix(test_y, y_pred)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xyjRehqeLDji",
        "outputId": "95ecf9c6-3d58-4591-b4b3-bf202b9e1373"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/40\n",
            "101/101 [==============================] - 19s 135ms/step - loss: 0.8235 - accuracy: 0.8147 - val_loss: 0.6309 - val_accuracy: 0.8151\n",
            "Epoch 2/40\n",
            "101/101 [==============================] - 7s 66ms/step - loss: 0.5864 - accuracy: 0.8224 - val_loss: 0.5851 - val_accuracy: 0.8151\n",
            "Epoch 3/40\n",
            "101/101 [==============================] - 7s 67ms/step - loss: 0.5453 - accuracy: 0.8225 - val_loss: 0.5472 - val_accuracy: 0.8151\n",
            "Epoch 4/40\n",
            "101/101 [==============================] - 5s 44ms/step - loss: 0.4965 - accuracy: 0.8269 - val_loss: 0.4909 - val_accuracy: 0.8283\n",
            "Epoch 5/40\n",
            "101/101 [==============================] - 5s 46ms/step - loss: 0.4357 - accuracy: 0.8472 - val_loss: 0.4456 - val_accuracy: 0.8413\n",
            "Epoch 6/40\n",
            "101/101 [==============================] - 4s 36ms/step - loss: 0.3915 - accuracy: 0.8597 - val_loss: 0.4245 - val_accuracy: 0.8443\n",
            "Epoch 7/40\n",
            "101/101 [==============================] - 4s 36ms/step - loss: 0.3626 - accuracy: 0.8685 - val_loss: 0.4140 - val_accuracy: 0.8443\n",
            "Epoch 8/40\n",
            "101/101 [==============================] - 3s 34ms/step - loss: 0.3355 - accuracy: 0.8780 - val_loss: 0.4124 - val_accuracy: 0.8471\n",
            "Epoch 9/40\n",
            "101/101 [==============================] - 4s 39ms/step - loss: 0.3092 - accuracy: 0.8858 - val_loss: 0.4045 - val_accuracy: 0.8535\n",
            "Epoch 10/40\n",
            "101/101 [==============================] - 3s 33ms/step - loss: 0.2883 - accuracy: 0.8908 - val_loss: 0.4050 - val_accuracy: 0.8532\n",
            "Epoch 11/40\n",
            "101/101 [==============================] - 4s 36ms/step - loss: 0.2644 - accuracy: 0.8999 - val_loss: 0.4206 - val_accuracy: 0.8508\n",
            "Epoch 12/40\n",
            "101/101 [==============================] - 3s 30ms/step - loss: 0.2513 - accuracy: 0.9030 - val_loss: 0.4366 - val_accuracy: 0.8535\n",
            "Epoch 13/40\n",
            "101/101 [==============================] - 4s 35ms/step - loss: 0.2366 - accuracy: 0.9091 - val_loss: 0.4374 - val_accuracy: 0.8542\n",
            "Epoch 14/40\n",
            "101/101 [==============================] - 3s 32ms/step - loss: 0.2238 - accuracy: 0.9133 - val_loss: 0.4433 - val_accuracy: 0.8535\n",
            "Epoch 15/40\n",
            "101/101 [==============================] - 3s 29ms/step - loss: 0.2127 - accuracy: 0.9187 - val_loss: 0.4531 - val_accuracy: 0.8576\n",
            "Epoch 16/40\n",
            "101/101 [==============================] - 3s 31ms/step - loss: 0.2049 - accuracy: 0.9212 - val_loss: 0.4627 - val_accuracy: 0.8542\n",
            "Epoch 17/40\n",
            "101/101 [==============================] - 3s 32ms/step - loss: 0.1965 - accuracy: 0.9238 - val_loss: 0.4786 - val_accuracy: 0.8563\n",
            "Epoch 18/40\n",
            "101/101 [==============================] - 3s 29ms/step - loss: 0.1872 - accuracy: 0.9276 - val_loss: 0.4730 - val_accuracy: 0.8573\n",
            "Epoch 19/40\n",
            "101/101 [==============================] - 3s 28ms/step - loss: 0.1818 - accuracy: 0.9316 - val_loss: 0.4918 - val_accuracy: 0.8559\n",
            "Epoch 20/40\n",
            "101/101 [==============================] - 3s 29ms/step - loss: 0.1732 - accuracy: 0.9357 - val_loss: 0.4899 - val_accuracy: 0.8532\n",
            "Epoch 21/40\n",
            "101/101 [==============================] - 3s 31ms/step - loss: 0.1661 - accuracy: 0.9377 - val_loss: 0.4970 - val_accuracy: 0.8518\n",
            "Epoch 22/40\n",
            "101/101 [==============================] - 3s 29ms/step - loss: 0.1581 - accuracy: 0.9424 - val_loss: 0.5049 - val_accuracy: 0.8542\n",
            "Epoch 23/40\n",
            "101/101 [==============================] - 3s 28ms/step - loss: 0.1528 - accuracy: 0.9454 - val_loss: 0.5105 - val_accuracy: 0.8484\n",
            "Epoch 24/40\n",
            "101/101 [==============================] - 3s 29ms/step - loss: 0.1463 - accuracy: 0.9472 - val_loss: 0.5298 - val_accuracy: 0.8539\n",
            "Epoch 25/40\n",
            "101/101 [==============================] - 3s 30ms/step - loss: 0.1404 - accuracy: 0.9496 - val_loss: 0.5267 - val_accuracy: 0.8559\n",
            "Epoch 26/40\n",
            "101/101 [==============================] - 3s 30ms/step - loss: 0.1351 - accuracy: 0.9525 - val_loss: 0.5539 - val_accuracy: 0.8532\n",
            "Epoch 27/40\n",
            "101/101 [==============================] - 3s 28ms/step - loss: 0.1298 - accuracy: 0.9549 - val_loss: 0.5389 - val_accuracy: 0.8559\n",
            "Epoch 28/40\n",
            "101/101 [==============================] - 3s 28ms/step - loss: 0.1215 - accuracy: 0.9568 - val_loss: 0.5466 - val_accuracy: 0.8576\n",
            "Epoch 29/40\n",
            "101/101 [==============================] - 3s 29ms/step - loss: 0.1158 - accuracy: 0.9594 - val_loss: 0.5566 - val_accuracy: 0.8542\n",
            "Epoch 30/40\n",
            "101/101 [==============================] - 3s 31ms/step - loss: 0.1113 - accuracy: 0.9616 - val_loss: 0.6079 - val_accuracy: 0.8430\n",
            "Epoch 31/40\n",
            "101/101 [==============================] - 3s 29ms/step - loss: 0.1083 - accuracy: 0.9624 - val_loss: 0.5534 - val_accuracy: 0.8426\n",
            "Epoch 32/40\n",
            "101/101 [==============================] - 3s 27ms/step - loss: 0.1059 - accuracy: 0.9629 - val_loss: 0.5831 - val_accuracy: 0.8535\n",
            "Epoch 33/40\n",
            "101/101 [==============================] - 3s 29ms/step - loss: 0.0990 - accuracy: 0.9662 - val_loss: 0.5808 - val_accuracy: 0.8535\n",
            "Epoch 34/40\n",
            "101/101 [==============================] - 3s 28ms/step - loss: 0.0969 - accuracy: 0.9663 - val_loss: 0.5840 - val_accuracy: 0.8495\n",
            "Epoch 35/40\n",
            "101/101 [==============================] - 3s 30ms/step - loss: 0.0932 - accuracy: 0.9680 - val_loss: 0.6006 - val_accuracy: 0.8532\n",
            "Epoch 36/40\n",
            "101/101 [==============================] - 3s 29ms/step - loss: 0.0898 - accuracy: 0.9684 - val_loss: 0.6027 - val_accuracy: 0.8392\n",
            "Epoch 37/40\n",
            "101/101 [==============================] - 3s 27ms/step - loss: 0.0898 - accuracy: 0.9691 - val_loss: 0.6098 - val_accuracy: 0.8467\n",
            "Epoch 38/40\n",
            "101/101 [==============================] - 3s 29ms/step - loss: 0.0870 - accuracy: 0.9707 - val_loss: 0.5980 - val_accuracy: 0.8539\n",
            "Epoch 39/40\n",
            "101/101 [==============================] - 3s 33ms/step - loss: 0.0841 - accuracy: 0.9716 - val_loss: 0.6011 - val_accuracy: 0.8454\n",
            "Epoch 40/40\n",
            "101/101 [==============================] - 3s 32ms/step - loss: 0.0817 - accuracy: 0.9718 - val_loss: 0.6442 - val_accuracy: 0.8457\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.10/dist-packages/keras/src/engine/training.py:3103: UserWarning: You are saving your model as an HDF5 file via `model.save()`. This file format is considered legacy. We recommend using instead the native Keras format, e.g. `model.save('my_model.keras')`.\n",
            "  saving_api.save_model(\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "f1_scores_per_class = f1_score(test_y, y_pred, average=None)\n",
        "\n",
        "# In F1 score cho mỗi class\n",
        "for i, score in enumerate(f1_scores_per_class):\n",
        "    print(f\"F1 Score for class {i}: {score}\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "q7hy1_Nr8Il6",
        "outputId": "bce0bf8d-923d-4715-e61c-9193644c8b35"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "F1 Score for class 0: 0.9239903001923239\n",
            "F1 Score for class 1: 0.3820459290187892\n",
            "F1 Score for class 2: 0.48726772195457674\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "accuracy_score(test_y, y_pred)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "yLRgjn7cMWdE",
        "outputId": "5927bfdb-0f1f-49ba-9589-40ab59a02e04"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.843562978427279"
            ]
          },
          "metadata": {},
          "execution_count": 47
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "f1_score(test_y, y_pred, average='micro')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "BjRA3L82MS8A",
        "outputId": "600199a1-7079-463f-b314-433d7b936617"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.843562978427279"
            ]
          },
          "metadata": {},
          "execution_count": 48
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "f1_score(test_y, y_pred, average='macro')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GnDvFpiYMTmx",
        "outputId": "5adf11d5-e7a9-4a42-f0d9-877e32b974c1"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.6013800148529169"
            ]
          },
          "metadata": {},
          "execution_count": 49
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "df_cm2 = pd.DataFrame(cf2, index = [\"normal\",\"hate\",\"harassment\"],\n",
        "                  columns = [\"normal\",\"hate\",\"harassment\"])\n",
        "plt.clf()\n",
        "sn.heatmap(df_cm2, annot=True, cmap=\"Greys\",fmt='g', cbar=True, annot_kws={\"size\": 30})"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 447
        },
        "id": "IN28uGkeQ5uV",
        "outputId": "093d7460-2c7a-4248-dd7c-633cba7adf98"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<Axes: >"
            ]
          },
          "metadata": {},
          "execution_count": 50
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Danh gia tren tap 2k7"
      ],
      "metadata": {
        "id": "a87bYzvPSKmj"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "prediction_2k7 = model.predict(test_X_2k7, batch_size=batch_size, verbose=0)\n",
        "y_pred_2k7 = prediction_2k7.argmax(axis=-1)\n",
        "\n",
        "cf2k7 = confusion_matrix(test_y_2k7, y_pred_2k7)"
      ],
      "metadata": {
        "id": "wVWYEgWuPVUE"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "accuracy_score(test_y_2k7, y_pred_2k7)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "iqUJqmELPsmB",
        "outputId": "b0c527b0-884a-4df7-f825-818c512f6f51"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.812962962962963"
            ]
          },
          "metadata": {},
          "execution_count": 52
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "f1_score(test_y_2k7, y_pred_2k7, average='macro')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "duS_mOQbQvsG",
        "outputId": "79189051-66a5-45b5-c256-b7e30bb73150"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.590143571417376"
            ]
          },
          "metadata": {},
          "execution_count": 53
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "df_cm2k7 = pd.DataFrame(cf2k7, index = [\"normal\",\"hate\",\"harassment\"],\n",
        "                  columns = [\"normal\",\"hate\",\"harassment\"])\n",
        "plt.clf()\n",
        "sn.heatmap(df_cm2k7, annot=True, cmap=\"Greys\",fmt='g', cbar=True, annot_kws={\"size\": 30})"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 447
        },
        "id": "ClpRnheQSQrd",
        "outputId": "682fe17c-7111-4d0c-8c25-5e200904ee0b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<Axes: >"
            ]
          },
          "metadata": {},
          "execution_count": 54
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## PhoBERT"
      ],
      "metadata": {
        "id": "YIF8U-4j-gcY"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!pip install vncorenlp"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "roOckqY0_FED",
        "outputId": "dabb8259-3ff7-444d-c423-984be818e35e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Requirement already satisfied: vncorenlp in /usr/local/lib/python3.10/dist-packages (1.0.3)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.10/dist-packages (from vncorenlp) (2.31.0)\n",
            "Requirement already satisfied: charset-normalizer<4,>=2 in /usr/local/lib/python3.10/dist-packages (from requests->vncorenlp) (3.3.2)\n",
            "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.10/dist-packages (from requests->vncorenlp) (3.6)\n",
            "Requirement already satisfied: urllib3<3,>=1.21.1 in /usr/local/lib/python3.10/dist-packages (from requests->vncorenlp) (2.0.7)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.10/dist-packages (from requests->vncorenlp) (2024.2.2)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "!mkdir -p vncorenlp/models/wordsegmenter\n",
        "!wget https://raw.githubusercontent.com/vncorenlp/VnCoreNLP/master/VnCoreNLP-1.1.1.jar\n",
        "!wget https://raw.githubusercontent.com/vncorenlp/VnCoreNLP/master/models/wordsegmenter/vi-vocab\n",
        "!wget https://raw.githubusercontent.com/vncorenlp/VnCoreNLP/master/models/wordsegmenter/wordsegmenter.rdr\n",
        "!mv VnCoreNLP-1.1.1.jar vncorenlp/\n",
        "!mv vi-vocab vncorenlp/models/wordsegmenter/\n",
        "!mv wordsegmenter.rdr vncorenlp/models/wordsegmenter/"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "P3L16VZB_Y88",
        "outputId": "a2ef92f0-d3d1-49a6-d190-7abe80233089"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "--2024-04-05 09:45:24--  https://raw.githubusercontent.com/vncorenlp/VnCoreNLP/master/VnCoreNLP-1.1.1.jar\n",
            "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.110.133, 185.199.111.133, 185.199.108.133, ...\n",
            "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.110.133|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 27412575 (26M) [application/octet-stream]\n",
            "Saving to: ‘VnCoreNLP-1.1.1.jar’\n",
            "\n",
            "VnCoreNLP-1.1.1.jar 100%[===================>]  26.14M  91.2MB/s    in 0.3s    \n",
            "\n",
            "2024-04-05 09:45:25 (91.2 MB/s) - ‘VnCoreNLP-1.1.1.jar’ saved [27412575/27412575]\n",
            "\n",
            "--2024-04-05 09:45:25--  https://raw.githubusercontent.com/vncorenlp/VnCoreNLP/master/models/wordsegmenter/vi-vocab\n",
            "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.108.133, 185.199.109.133, 185.199.110.133, ...\n",
            "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.108.133|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 526544 (514K) [application/octet-stream]\n",
            "Saving to: ‘vi-vocab’\n",
            "\n",
            "vi-vocab            100%[===================>] 514.20K  --.-KB/s    in 0.04s   \n",
            "\n",
            "2024-04-05 09:45:25 (13.5 MB/s) - ‘vi-vocab’ saved [526544/526544]\n",
            "\n",
            "--2024-04-05 09:45:25--  https://raw.githubusercontent.com/vncorenlp/VnCoreNLP/master/models/wordsegmenter/wordsegmenter.rdr\n",
            "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.108.133, 185.199.109.133, 185.199.110.133, ...\n",
            "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.108.133|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 128508 (125K) [text/plain]\n",
            "Saving to: ‘wordsegmenter.rdr’\n",
            "\n",
            "wordsegmenter.rdr   100%[===================>] 125.50K  --.-KB/s    in 0.02s   \n",
            "\n",
            "2024-04-05 09:45:25 (5.91 MB/s) - ‘wordsegmenter.rdr’ saved [128508/128508]\n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from vncorenlp import VnCoreNLP\n",
        "\n",
        "vncorenlp = VnCoreNLP(\"vncorenlp/VnCoreNLP-1.1.1.jar\", annotators=\"wseg\", max_heap_size='-Xmx500m')"
      ],
      "metadata": {
        "id": "ejrUobtI-3LR"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def preprocess(text, tokenized = True, lowercased = True):\n",
        "    # text = filter_stop_words(text, stopwords)\n",
        "    # text = remove_special_characters(text)\n",
        "    # text = emoji.replace_emoji(text,\"\")\n",
        "    text = text.lower() if lowercased else text\n",
        "    if tokenized:\n",
        "        pre_text = \"\"\n",
        "        sentences = vncorenlp.tokenize(text)\n",
        "        for sentence in sentences:\n",
        "            pre_text += \" \".join(sentence)\n",
        "        text = pre_text\n",
        "    return text\n",
        "\n",
        "def pre_process_features(X, y, tokenized=True, lowercased=True):\n",
        "    X_processed = [preprocess(str(p), tokenized=tokenized, lowercased=lowercased) for p in list(X)]\n",
        "    non_empty_indices = [idx for idx, ele in enumerate(X_processed) if ele]\n",
        "    X_processed = pd.Series(X_processed)[non_empty_indices].tolist()\n",
        "    y_processed = y.iloc[non_empty_indices].tolist()\n",
        "    return X_processed, y_processed"
      ],
      "metadata": {
        "id": "g-4aF7vl_80t"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "train_X, train_y = pre_process_features(combined_train['text'], combined_train['label_id'], tokenized=True, lowercased = False)\n",
        "dev_X, dev_y = pre_process_features(combined_dev['text'], combined_dev['label_id'], tokenized=True, lowercased = False)\n",
        "test_X, test_y = pre_process_features(combined_test['text'], combined_test['label_id'], tokenized=True, lowercased = False)"
      ],
      "metadata": {
        "id": "XVX7W6IHDuRe"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "test_X_2k7, test_y_2k7 = pre_process_features(test['text'], test['label'], tokenized=True, lowercased = False)"
      ],
      "metadata": {
        "id": "LL7rO4dgbg8G"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "pip install transformers[torch] accelerate -U"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "nhdcKyFOOU2e",
        "outputId": "b4582c91-92ef-42f6-b023-89633a827834"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Requirement already satisfied: transformers[torch] in /usr/local/lib/python3.10/dist-packages (4.38.2)\n",
            "Collecting transformers[torch]\n",
            "  Downloading transformers-4.39.3-py3-none-any.whl (8.8 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m8.8/8.8 MB\u001b[0m \u001b[31m15.6 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting accelerate\n",
            "  Downloading accelerate-0.28.0-py3-none-any.whl (290 kB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m290.1/290.1 kB\u001b[0m \u001b[31m6.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: filelock in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (3.13.3)\n",
            "Requirement already satisfied: huggingface-hub<1.0,>=0.19.3 in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (0.20.3)\n",
            "Requirement already satisfied: numpy>=1.17 in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (1.25.2)\n",
            "Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (24.0)\n",
            "Requirement already satisfied: pyyaml>=5.1 in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (6.0.1)\n",
            "Requirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (2023.12.25)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (2.31.0)\n",
            "Requirement already satisfied: tokenizers<0.19,>=0.14 in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (0.15.2)\n",
            "Requirement already satisfied: safetensors>=0.4.1 in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (0.4.2)\n",
            "Requirement already satisfied: tqdm>=4.27 in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (4.66.2)\n",
            "Requirement already satisfied: torch in /usr/local/lib/python3.10/dist-packages (from transformers[torch]) (2.2.1+cu121)\n",
            "Requirement already satisfied: psutil in /usr/local/lib/python3.10/dist-packages (from accelerate) (5.9.5)\n",
            "Requirement already satisfied: fsspec>=2023.5.0 in /usr/local/lib/python3.10/dist-packages (from huggingface-hub<1.0,>=0.19.3->transformers[torch]) (2023.6.0)\n",
            "Requirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.10/dist-packages (from huggingface-hub<1.0,>=0.19.3->transformers[torch]) (4.10.0)\n",
            "Requirement already satisfied: sympy in /usr/local/lib/python3.10/dist-packages (from torch->transformers[torch]) (1.12)\n",
            "Requirement already satisfied: networkx in /usr/local/lib/python3.10/dist-packages (from torch->transformers[torch]) (3.2.1)\n",
            "Requirement already satisfied: jinja2 in /usr/local/lib/python3.10/dist-packages (from torch->transformers[torch]) (3.1.3)\n",
            "Collecting nvidia-cuda-nvrtc-cu12==12.1.105 (from torch->transformers[torch])\n",
            "  Downloading nvidia_cuda_nvrtc_cu12-12.1.105-py3-none-manylinux1_x86_64.whl (23.7 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m23.7/23.7 MB\u001b[0m \u001b[31m24.9 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting nvidia-cuda-runtime-cu12==12.1.105 (from torch->transformers[torch])\n",
            "  Downloading nvidia_cuda_runtime_cu12-12.1.105-py3-none-manylinux1_x86_64.whl (823 kB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m823.6/823.6 kB\u001b[0m \u001b[31m10.9 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting nvidia-cuda-cupti-cu12==12.1.105 (from torch->transformers[torch])\n",
            "  Downloading nvidia_cuda_cupti_cu12-12.1.105-py3-none-manylinux1_x86_64.whl (14.1 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m14.1/14.1 MB\u001b[0m \u001b[31m34.0 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting nvidia-cudnn-cu12==8.9.2.26 (from torch->transformers[torch])\n",
            "  Downloading nvidia_cudnn_cu12-8.9.2.26-py3-none-manylinux1_x86_64.whl (731.7 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m731.7/731.7 MB\u001b[0m \u001b[31m2.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting nvidia-cublas-cu12==12.1.3.1 (from torch->transformers[torch])\n",
            "  Downloading nvidia_cublas_cu12-12.1.3.1-py3-none-manylinux1_x86_64.whl (410.6 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m410.6/410.6 MB\u001b[0m \u001b[31m3.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting nvidia-cufft-cu12==11.0.2.54 (from torch->transformers[torch])\n",
            "  Downloading nvidia_cufft_cu12-11.0.2.54-py3-none-manylinux1_x86_64.whl (121.6 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m121.6/121.6 MB\u001b[0m \u001b[31m6.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting nvidia-curand-cu12==10.3.2.106 (from torch->transformers[torch])\n",
            "  Downloading nvidia_curand_cu12-10.3.2.106-py3-none-manylinux1_x86_64.whl (56.5 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m56.5/56.5 MB\u001b[0m \u001b[31m9.0 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting nvidia-cusolver-cu12==11.4.5.107 (from torch->transformers[torch])\n",
            "  Downloading nvidia_cusolver_cu12-11.4.5.107-py3-none-manylinux1_x86_64.whl (124.2 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m124.2/124.2 MB\u001b[0m \u001b[31m2.9 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting nvidia-cusparse-cu12==12.1.0.106 (from torch->transformers[torch])\n",
            "  Downloading nvidia_cusparse_cu12-12.1.0.106-py3-none-manylinux1_x86_64.whl (196.0 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m196.0/196.0 MB\u001b[0m \u001b[31m3.3 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting nvidia-nccl-cu12==2.19.3 (from torch->transformers[torch])\n",
            "  Downloading nvidia_nccl_cu12-2.19.3-py3-none-manylinux1_x86_64.whl (166.0 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m166.0/166.0 MB\u001b[0m \u001b[31m3.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hCollecting nvidia-nvtx-cu12==12.1.105 (from torch->transformers[torch])\n",
            "  Downloading nvidia_nvtx_cu12-12.1.105-py3-none-manylinux1_x86_64.whl (99 kB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m99.1/99.1 kB\u001b[0m \u001b[31m5.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: triton==2.2.0 in /usr/local/lib/python3.10/dist-packages (from torch->transformers[torch]) (2.2.0)\n",
            "Collecting nvidia-nvjitlink-cu12 (from nvidia-cusolver-cu12==11.4.5.107->torch->transformers[torch])\n",
            "  Downloading nvidia_nvjitlink_cu12-12.4.127-py3-none-manylinux2014_x86_64.whl (21.1 MB)\n",
            "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m21.1/21.1 MB\u001b[0m \u001b[31m4.5 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
            "\u001b[?25hRequirement already satisfied: charset-normalizer<4,>=2 in /usr/local/lib/python3.10/dist-packages (from requests->transformers[torch]) (3.3.2)\n",
            "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.10/dist-packages (from requests->transformers[torch]) (3.6)\n",
            "Requirement already satisfied: urllib3<3,>=1.21.1 in /usr/local/lib/python3.10/dist-packages (from requests->transformers[torch]) (2.0.7)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.10/dist-packages (from requests->transformers[torch]) (2024.2.2)\n",
            "Requirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.10/dist-packages (from jinja2->torch->transformers[torch]) (2.1.5)\n",
            "Requirement already satisfied: mpmath>=0.19 in /usr/local/lib/python3.10/dist-packages (from sympy->torch->transformers[torch]) (1.3.0)\n",
            "Installing collected packages: nvidia-nvtx-cu12, nvidia-nvjitlink-cu12, nvidia-nccl-cu12, nvidia-curand-cu12, nvidia-cufft-cu12, nvidia-cuda-runtime-cu12, nvidia-cuda-nvrtc-cu12, nvidia-cuda-cupti-cu12, nvidia-cublas-cu12, nvidia-cusparse-cu12, nvidia-cudnn-cu12, nvidia-cusolver-cu12, transformers, accelerate\n",
            "  Attempting uninstall: transformers\n",
            "    Found existing installation: transformers 4.38.2\n",
            "    Uninstalling transformers-4.38.2:\n",
            "      Successfully uninstalled transformers-4.38.2\n",
            "Successfully installed accelerate-0.28.0 nvidia-cublas-cu12-12.1.3.1 nvidia-cuda-cupti-cu12-12.1.105 nvidia-cuda-nvrtc-cu12-12.1.105 nvidia-cuda-runtime-cu12-12.1.105 nvidia-cudnn-cu12-8.9.2.26 nvidia-cufft-cu12-11.0.2.54 nvidia-curand-cu12-10.3.2.106 nvidia-cusolver-cu12-11.4.5.107 nvidia-cusparse-cu12-12.1.0.106 nvidia-nccl-cu12-2.19.3 nvidia-nvjitlink-cu12-12.4.127 nvidia-nvtx-cu12-12.1.105 transformers-4.39.3\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from transformers import AutoTokenizer, AutoModelForSequenceClassification, Trainer, TrainingArguments, BertTokenizer, BertForSequenceClassification"
      ],
      "metadata": {
        "id": "yXgeFvRiHGt_"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model = AutoModelForSequenceClassification.from_pretrained(\"vinai/phobert-base\", num_labels = 3)\n",
        "tokenizer = AutoTokenizer.from_pretrained(\"vinai/phobert-base\",use_fast=False)\n",
        "\n",
        "import torch\n",
        "\n",
        "class BuildDataset(torch.utils.data.Dataset):\n",
        "    def __init__(self, encodings, labels):\n",
        "        self.encodings = encodings\n",
        "        self.labels = labels\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}\n",
        "        item['labels'] = torch.tensor(self.labels[idx])\n",
        "        return item\n",
        "\n",
        "    def __len__(self):\n",
        "        return len(self.labels)\n",
        "\n",
        "train_encodings = tokenizer(train_X, truncation=True, padding=True, max_length=100)\n",
        "dev_encodings = tokenizer(dev_X, truncation=True, padding=True, max_length=100)\n",
        "test_encodings = tokenizer(test_X, truncation=True, padding=True, max_length=100)\n",
        "test_2k7_encodings = tokenizer(test_X_2k7, truncation=True, padding=True, max_length=100)\n",
        "\n",
        "train_dataset = BuildDataset(train_encodings, train_y)\n",
        "dev_dataset = BuildDataset(dev_encodings, dev_y)\n",
        "test_dataset = BuildDataset(test_encodings, test_y)\n",
        "test_2k7 = BuildDataset(test_2k7_encodings, test_y_2k7)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7A36e9JTG-4m",
        "outputId": "f9e41a6c-102c-46c2-c803-39bd350d7c63"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Some weights of RobertaForSequenceClassification were not initialized from the model checkpoint at vinai/phobert-base and are newly initialized: ['classifier.dense.bias', 'classifier.dense.weight', 'classifier.out_proj.bias', 'classifier.out_proj.weight']\n",
            "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "training_args = TrainingArguments(\n",
        "    output_dir=DATA_PATH + \"transformer_model/phobert-v3/\",\n",
        "    num_train_epochs=2,\n",
        "    per_device_train_batch_size=16,\n",
        "    per_device_eval_batch_size=16,\n",
        "    warmup_steps=500,\n",
        "    weight_decay=0.01,\n",
        "    no_cuda=False,\n",
        "    do_eval=True\n",
        ")\n",
        "\n",
        "trainer = Trainer(\n",
        "    model=model,\n",
        "    args=training_args,\n",
        "    train_dataset=train_dataset,\n",
        "    eval_dataset=dev_dataset\n",
        ")\n",
        "\n",
        "trainer.train()\n",
        "trainer.save_model(DATA_PATH + \"/transformer_model/phobert-v3/\")\n",
        "\n",
        "# y_pred_classify = trainer.predict(test_dataset)\n",
        "\n",
        "# y_pred = np.argmax(y_pred_classify.predictions, axis=-1)\n",
        "# y_true = test_y"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 335
        },
        "id": "lBDKYSd8IC74",
        "outputId": "d869ac93-2a63-4f40-bf4a-69df2d5a7730"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.10/dist-packages/accelerate/accelerator.py:432: FutureWarning: Passing the following arguments to `Accelerator` is deprecated and will be removed in version 1.0 of Accelerate: dict_keys(['dispatch_batches', 'split_batches', 'even_batches', 'use_seedable_sampler']). Please pass an `accelerate.DataLoaderConfiguration` instead: \n",
            "dataloader_config = DataLoaderConfiguration(dispatch_batches=None, split_batches=False, even_batches=True, use_seedable_sampler=True)\n",
            "  warnings.warn(\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ],
            "text/html": [
              "\n",
              "    <div>\n",
              "      \n",
              "      <progress value='3242' max='3242' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      [3242/3242 18:25, Epoch 2/2]\n",
              "    </div>\n",
              "    <table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              " <tr style=\"text-align: left;\">\n",
              "      <th>Step</th>\n",
              "      <th>Training Loss</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>500</td>\n",
              "      <td>0.555100</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1000</td>\n",
              "      <td>0.434700</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1500</td>\n",
              "      <td>0.394900</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2000</td>\n",
              "      <td>0.326600</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2500</td>\n",
              "      <td>0.289600</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3000</td>\n",
              "      <td>0.283300</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table><p>"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ],
            "text/html": []
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "model = AutoModelForSequenceClassification.from_pretrained(DATA_PATH + \"/transformer_model/phobert-v3/\")"
      ],
      "metadata": {
        "id": "pvoPzc5hhTNw"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "training_args = TrainingArguments(\n",
        "    output_dir=DATA_PATH + \"transformer_model/phobert-v3/\",\n",
        "    num_train_epochs=2,\n",
        "    per_device_train_batch_size=16,\n",
        "    per_device_eval_batch_size=16,\n",
        "    warmup_steps=500,\n",
        "    weight_decay=0.01,\n",
        "    no_cuda=False,\n",
        "    do_eval=True\n",
        ")\n",
        "trainer = Trainer(\n",
        "    model=model,\n",
        "    args=training_args,\n",
        "    train_dataset=train_dataset,\n",
        "    eval_dataset=dev_dataset\n",
        ")\n",
        "\n",
        "# Predict on the test dataset\n",
        "y_pred_classify = trainer.predict(test_dataset)\n",
        "\n",
        "# Extract predictions\n",
        "y_pred = np.argmax(y_pred_classify.predictions, axis=-1)\n",
        "y_true = test_y"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 89
        },
        "id": "WQWrirXHlNFW",
        "outputId": "043abb97-136e-46b3-ee60-3a7754e052c9"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.10/dist-packages/accelerate/accelerator.py:432: FutureWarning: Passing the following arguments to `Accelerator` is deprecated and will be removed in version 1.0 of Accelerate: dict_keys(['dispatch_batches', 'split_batches', 'even_batches', 'use_seedable_sampler']). Please pass an `accelerate.DataLoaderConfiguration` instead: \n",
            "dataloader_config = DataLoaderConfiguration(dispatch_batches=None, split_batches=False, even_batches=True, use_seedable_sampler=True)\n",
            "  warnings.warn(\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ],
            "text/html": []
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "cf1 = confusion_matrix(y_true, y_pred)"
      ],
      "metadata": {
        "id": "PQaVRROyWU1U"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "f1_scores_per_class = f1_score(y_true, y_pred, average=None)\n",
        "\n",
        "# In F1 score cho mỗi class\n",
        "for i, score in enumerate(f1_scores_per_class):\n",
        "    print(f\"F1 Score for class {i}: {score}\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "SIT4vKdR8Zb0",
        "outputId": "628cfc8a-2088-48db-bd7c-e805c776a8cc"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "F1 Score for class 0: 0.9408668476025989\n",
            "F1 Score for class 1: 0.43394199785177234\n",
            "F1 Score for class 2: 0.5792592592592593\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "accuracy_score(y_true,y_pred)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "atHFK9aaWlws",
        "outputId": "6d366c07-2778-49b8-8a9e-d2c3d8eba113"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.874376731301939"
            ]
          },
          "metadata": {},
          "execution_count": 74
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "f1_score(y_true, y_pred, average='micro')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "PMDQOssZWY0S",
        "outputId": "ac59707b-9d77-43a3-9ce2-a3411f47485a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.874376731301939"
            ]
          },
          "metadata": {},
          "execution_count": 66
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "f1_score(y_true, y_pred, average='macro')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "MW2K2hHvWc0i",
        "outputId": "b4e54522-1d84-49be-a851-56a538be3ce3"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.6513560349045435"
            ]
          },
          "metadata": {},
          "execution_count": 67
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "df_cm1 = pd.DataFrame(cf1, index = [\"normal\",\"hate\",\"harassment\"],\n",
        "                  columns = [\"normal\",\"hate\",\"harassment\"])\n",
        "plt.clf()\n",
        "sn.heatmap(df_cm1, annot=True, cmap=\"Greys\",fmt='g', cbar=True, annot_kws={\"size\": 30})"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 447
        },
        "id": "8mymdaU2Whei",
        "outputId": "deaec4ce-b4ab-4bb8-fb89-3d1dbb5b396d"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<Axes: >"
            ]
          },
          "metadata": {},
          "execution_count": 75
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "prediction_2k7 = trainer.predict(test_2k7)\n",
        "y_pred_2k7 = np.argmax(prediction_2k7.predictions, axis=-1)\n",
        "\n",
        "cf2k7 = confusion_matrix(test_y_2k7, y_pred_2k7)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 17
        },
        "id": "bYtt4DeIaCel",
        "outputId": "98f52e62-9716-40f5-e576-afac6de6212c"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ],
            "text/html": []
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "accuracy_score(test_y_2k7, y_pred_2k7)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "o4mfqEqidbUx",
        "outputId": "82174996-0692-46a0-a1b4-69fbf5a2d6a5"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.8648148148148148"
            ]
          },
          "metadata": {},
          "execution_count": 70
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "f1_score(test_y_2k7, y_pred_2k7, average='macro')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CEbOCbg1dh2g",
        "outputId": "e1cf955b-ed1d-4ff8-c873-166b4d8860cb"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.6791154424281035"
            ]
          },
          "metadata": {},
          "execution_count": 71
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "df_cm2k7 = pd.DataFrame(cf2k7, index = [\"normal\",\"hate\",\"harassment\"],\n",
        "                  columns = [\"normal\",\"hate\",\"harassment\"])\n",
        "plt.clf()\n",
        "sn.heatmap(df_cm2k7, annot=True, cmap=\"Greys\",fmt='g', cbar=True, annot_kws={\"size\": 30})"
      ],
      "metadata": {
        "id": "5FZXm2YXdms9",
        "outputId": "7c36f010-ccda-4c72-c400-519f3c91c26b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 450
        }
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<Axes: >"
            ]
          },
          "metadata": {},
          "execution_count": 72
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "misclassified_indices = np.where((np.array(y_true) != 0) & (y_pred == 0))[0]"
      ],
      "metadata": {
        "id": "KtBaiJHGcA9t"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "len(misclassified_indices)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "btgK0JcgcDix",
        "outputId": "214c01f5-80d1-4d96-f042-029dd30a28ad"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "474"
            ]
          },
          "metadata": {},
          "execution_count": 26
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "combined_test.iloc[misclassified_indices].head(20)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 676
        },
        "id": "vErSTD-_cXp1",
        "outputId": "86f146ea-9201-4a09-cb60-acc9d36a9003"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                                  text  label_id\n",
              "0    từ dịch này thấy ý thức người nhật xăng đổ nửa...         2\n",
              "14                 K có nghìn nào giúp dân thì nín đi!         2\n",
              "16    Yenie Cao riết mặp như heo r mà cho ăn quài 😭😭😭😭         2\n",
              "18                Nguyễn Đức Ánh khôn hơn cả họ nhà m.         2\n",
              "21                        Tây nó lại bảo mới trốn viện         1\n",
              "25                            Trùm cuối cmn hú vía :))         1\n",
              "82   Súc vật từ nhan sắc đến nhân cách súc phạm loà...         2\n",
              "111                                  Xem bực mình thêm         1\n",
              "121  Cuốn dell nhặt đc mồm, toàn danh hài đất Việt :))         1\n",
              "138          Chỗ này đit phải chỗ gạ địt nhé hahhahah.         1\n",
              "166  Phải,VN vô số tham nhũng,lại vụ nữa đây,sắp đạ...         1\n",
              "192  Bất hạnh nhất cuộc đời là ra trường rồi mà bạn...         1\n",
              "208  bị tâm thần vẫn có thể có học vị TS nh...         2\n",
              "214  A đừng có giống chị Tịp đấy nhé, thông báo các...         2\n",
              "263                                           loạn cực         1\n",
              "269                                   thầy chơi bẩn vl         2\n",
              "284                            Bú l ngọc Trinh sẽ tinh         2\n",
              "290  Mua sắt vụn về làm phương tiện chiến đấu! Đắng...         1\n",
              "299  Tao gõ chớt mẹ mày chứ ở đó mà múa . Abe said :))         2\n",
              "302  có tiền cứ làm,chỉ có bọn không có tiền làm mớ...         2"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-04e9cc3c-fdb5-45bd-bfa5-764bfb377289\" class=\"colab-df-container\">\n",
              "    <div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>text</th>\n",
              "      <th>label_id</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>từ dịch này thấy ý thức người nhật xăng đổ nửa...</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>14</th>\n",
              "      <td>K có nghìn nào giúp dân thì nín đi!</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>16</th>\n",
              "      <td>Yenie Cao riết mặp như heo r mà cho ăn quài 😭😭😭😭</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>18</th>\n",
              "      <td>Nguyễn Đức Ánh khôn hơn cả họ nhà m.</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>21</th>\n",
              "      <td>Tây nó lại bảo mới trốn viện</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>25</th>\n",
              "      <td>Trùm cuối cmn hú vía :))</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>82</th>\n",
              "      <td>Súc vật từ nhan sắc đến nhân cách súc phạm loà...</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>111</th>\n",
              "      <td>Xem bực mình thêm</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>121</th>\n",
              "      <td>Cuốn dell nhặt đc mồm, toàn danh hài đất Việt :))</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>138</th>\n",
              "      <td>Chỗ này đit phải chỗ gạ địt nhé hahhahah.</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>166</th>\n",
              "      <td>Phải,VN vô số tham nhũng,lại vụ nữa đây,sắp đạ...</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>192</th>\n",
              "      <td>Bất hạnh nhất cuộc đời là ra trường rồi mà bạn...</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>208</th>\n",
              "      <td>bị tâm thần vẫn có thể có học vị TS nh...</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>214</th>\n",
              "      <td>A đừng có giống chị Tịp đấy nhé, thông báo các...</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>263</th>\n",
              "      <td>loạn cực</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>269</th>\n",
              "      <td>thầy chơi bẩn vl</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>284</th>\n",
              "      <td>Bú l ngọc Trinh sẽ tinh</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>290</th>\n",
              "      <td>Mua sắt vụn về làm phương tiện chiến đấu! Đắng...</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>299</th>\n",
              "      <td>Tao gõ chớt mẹ mày chứ ở đó mà múa . Abe said :))</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>302</th>\n",
              "      <td>có tiền cứ làm,chỉ có bọn không có tiền làm mớ...</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "    <div class=\"colab-df-buttons\">\n",
              "\n",
              "  <div class=\"colab-df-container\">\n",
              "    <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-04e9cc3c-fdb5-45bd-bfa5-764bfb377289')\"\n",
              "            title=\"Convert this dataframe to an interactive table.\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\" viewBox=\"0 -960 960 960\">\n",
              "    <path d=\"M120-120v-720h720v720H120Zm60-500h600v-160H180v160Zm220 220h160v-160H400v160Zm0 220h160v-160H400v160ZM180-400h160v-160H180v160Zm440 0h160v-160H620v160ZM180-180h160v-160H180v160Zm440 0h160v-160H620v160Z\"/>\n",
              "  </svg>\n",
              "    </button>\n",
              "\n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    .colab-df-buttons div {\n",
              "      margin-bottom: 4px;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "    <script>\n",
              "      const buttonEl =\n",
              "        document.querySelector('#df-04e9cc3c-fdb5-45bd-bfa5-764bfb377289 button.colab-df-convert');\n",
              "      buttonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "      async function convertToInteractive(key) {\n",
              "        const element = document.querySelector('#df-04e9cc3c-fdb5-45bd-bfa5-764bfb377289');\n",
              "        const dataTable =\n",
              "          await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                    [key], {});\n",
              "        if (!dataTable) return;\n",
              "\n",
              "        const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "          '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "          + ' to learn more about interactive tables.';\n",
              "        element.innerHTML = '';\n",
              "        dataTable['output_type'] = 'display_data';\n",
              "        await google.colab.output.renderOutput(dataTable, element);\n",
              "        const docLink = document.createElement('div');\n",
              "        docLink.innerHTML = docLinkHtml;\n",
              "        element.appendChild(docLink);\n",
              "      }\n",
              "    </script>\n",
              "  </div>\n",
              "\n",
              "\n",
              "<div id=\"df-7918245a-2def-49b0-b5b7-675059633406\">\n",
              "  <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-7918245a-2def-49b0-b5b7-675059633406')\"\n",
              "            title=\"Suggest charts\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "<svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "     width=\"24px\">\n",
              "    <g>\n",
              "        <path d=\"M19 3H5c-1.1 0-2 .9-2 2v14c0 1.1.9 2 2 2h14c1.1 0 2-.9 2-2V5c0-1.1-.9-2-2-2zM9 17H7v-7h2v7zm4 0h-2V7h2v10zm4 0h-2v-4h2v4z\"/>\n",
              "    </g>\n",
              "</svg>\n",
              "  </button>\n",
              "\n",
              "<style>\n",
              "  .colab-df-quickchart {\n",
              "      --bg-color: #E8F0FE;\n",
              "      --fill-color: #1967D2;\n",
              "      --hover-bg-color: #E2EBFA;\n",
              "      --hover-fill-color: #174EA6;\n",
              "      --disabled-fill-color: #AAA;\n",
              "      --disabled-bg-color: #DDD;\n",
              "  }\n",
              "\n",
              "  [theme=dark] .colab-df-quickchart {\n",
              "      --bg-color: #3B4455;\n",
              "      --fill-color: #D2E3FC;\n",
              "      --hover-bg-color: #434B5C;\n",
              "      --hover-fill-color: #FFFFFF;\n",
              "      --disabled-bg-color: #3B4455;\n",
              "      --disabled-fill-color: #666;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart {\n",
              "    background-color: var(--bg-color);\n",
              "    border: none;\n",
              "    border-radius: 50%;\n",
              "    cursor: pointer;\n",
              "    display: none;\n",
              "    fill: var(--fill-color);\n",
              "    height: 32px;\n",
              "    padding: 0;\n",
              "    width: 32px;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart:hover {\n",
              "    background-color: var(--hover-bg-color);\n",
              "    box-shadow: 0 1px 2px rgba(60, 64, 67, 0.3), 0 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "    fill: var(--button-hover-fill-color);\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart-complete:disabled,\n",
              "  .colab-df-quickchart-complete:disabled:hover {\n",
              "    background-color: var(--disabled-bg-color);\n",
              "    fill: var(--disabled-fill-color);\n",
              "    box-shadow: none;\n",
              "  }\n",
              "\n",
              "  .colab-df-spinner {\n",
              "    border: 2px solid var(--fill-color);\n",
              "    border-color: transparent;\n",
              "    border-bottom-color: var(--fill-color);\n",
              "    animation:\n",
              "      spin 1s steps(1) infinite;\n",
              "  }\n",
              "\n",
              "  @keyframes spin {\n",
              "    0% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "      border-left-color: var(--fill-color);\n",
              "    }\n",
              "    20% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    30% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    40% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    60% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    80% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "    90% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "  }\n",
              "</style>\n",
              "\n",
              "  <script>\n",
              "    async function quickchart(key) {\n",
              "      const quickchartButtonEl =\n",
              "        document.querySelector('#' + key + ' button');\n",
              "      quickchartButtonEl.disabled = true;  // To prevent multiple clicks.\n",
              "      quickchartButtonEl.classList.add('colab-df-spinner');\n",
              "      try {\n",
              "        const charts = await google.colab.kernel.invokeFunction(\n",
              "            'suggestCharts', [key], {});\n",
              "      } catch (error) {\n",
              "        console.error('Error during call to suggestCharts:', error);\n",
              "      }\n",
              "      quickchartButtonEl.classList.remove('colab-df-spinner');\n",
              "      quickchartButtonEl.classList.add('colab-df-quickchart-complete');\n",
              "    }\n",
              "    (() => {\n",
              "      let quickchartButtonEl =\n",
              "        document.querySelector('#df-7918245a-2def-49b0-b5b7-675059633406 button');\n",
              "      quickchartButtonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "    })();\n",
              "  </script>\n",
              "</div>\n",
              "\n",
              "    </div>\n",
              "  </div>\n"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "dataframe",
              "summary": "{\n  \"name\": \"combined_test\",\n  \"rows\": 20,\n  \"fields\": [\n    {\n      \"column\": \"text\",\n      \"properties\": {\n        \"dtype\": \"string\",\n        \"num_unique_values\": 20,\n        \"samples\": [\n          \"t\\u1eeb d\\u1ecbch n\\u00e0y th\\u1ea5y \\u00fd th\\u1ee9c ng\\u01b0\\u1eddi nh\\u1eadt x\\u0103ng \\u0111\\u1ed5 n\\u1eeda b\\u00ecnh, nh\\u01b0ng gi\\u1ea5y v\\u1ec7 sinh kh\\u00f4ng th\\u1ec3 thi\\u1ebfu. kh\\u00f4ng hy sinh b\\u1ea3n th\\u00e2n \\u0111\\u1ec3 c\\u1ee9u ng\\u01b0\\u1eddi kh\\u00e1c, nh\\u01b0ng d\\u00ednh d\\u1ecbch th\\u00ec k\\u00e9o nhi\\u1ec1u ng\\u01b0\\u1eddi d\\u00ednh theo.\",\n          \"Mua s\\u1eaft v\\u1ee5n v\\u1ec1 l\\u00e0m ph\\u01b0\\u01a1ng ti\\u1ec7n chi\\u1ebfn \\u0111\\u1ea5u! \\u0110\\u1eafng l\\u00f2ng\",\n          \"th\\u1ea7y ch\\u01a1i b\\u1ea9n vl\"\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    },\n    {\n      \"column\": \"label_id\",\n      \"properties\": {\n        \"dtype\": \"number\",\n        \"std\": 0,\n        \"min\": 1,\n        \"max\": 2,\n        \"num_unique_values\": 2,\n        \"samples\": [\n          1,\n          2\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    }\n  ]\n}"
            }
          },
          "metadata": {},
          "execution_count": 28
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "misclassified_indices = np.where((np.array(test_y_2k7) ==1) & (y_pred_2k7 == 2))[0]"
      ],
      "metadata": {
        "id": "kb-pbzzYyMPQ"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "len(misclassified_indices)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_G4FYOb4yban",
        "outputId": "17c1165e-c163-456b-ddde-99c96a7bd20a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "32"
            ]
          },
          "metadata": {},
          "execution_count": 36
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "pd.set_option('max_colwidth', 800)"
      ],
      "metadata": {
        "id": "3ZhyzTFF0L9n"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "test.iloc[misclassified_indices]"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "3aFEImRDy41P",
        "outputId": "78983cbc-f234-444c-a423-c30507c40fed"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                                                                                                                                                                                                                                                                   text  label\n",
              "1162                                                                                                                                                                                                         Phạm Thái Hoàng Long ồ loại m đ cãi dc là lại địt mẹ mày à, loạn luân vcl       1\n",
              "161                                                                                                                                                                                                                                          Joseph Tuấn Anh T mà ăn hiếp m à thằng lồn      1\n",
              "1507                                                                                                                                                                                                            Anh Duy lại chó nữa🐶\\nCái dm chó cắn chó ăn cứt đi đừng có đi cắn người      1\n",
              "1920                                                                                                                                                                                                                       Khanh Pham Van mày đi hỏi nó đi! Đm ngu lồn tỏ vẻ hiểu biết.      1\n",
              "1395                                                                                                                                                                                                                         Má sử cho đẹp con đĩ đó nha má nghe nó ns mà tức thiệt chứ      1\n",
              "1211                                                                                                                                                                                                                               Hiền Hén thông cảm cho con mụ điên đấy tý đi chị êii      1\n",
              "2106                                                                                                                Tui xin hỏi tụi con gái thử mày bị đánh trc mặt biết bao người v có biết nhục ko may cho mày là kiếm đc thằng ck bt nhịn mày gặp máy đứa khác đánh cho mày tung lôl      1\n",
              "2229                                                                                                                                    Dcm may thang lon chung may ngu vay. Bon cho cong an ko đánh nhung no goi nguòi khac toi đánh. Chúng may song ma éo biet gi chet het đi bon ngu      1\n",
              "2161  Hùng Sơn thằng đéo bò chữa hoang này.mày biết tao vùng nào không mà vào sủa  vậy? Dòng họ tao điều có 1 số người làm việc và theo 2 chế độ.tôn giáo cũng có thiên chúa nghen thằng đầu bò.nhưng khi hòa bình 2 phía vẫn nhìn nhau và giúp đỡ lẫn nhau      1\n",
              "567                                           Hồng Thơm  hết chia sẻ rồi lại bình luận xàm lồn để lzi để chúng nó thấy rồi cười vào mặt cho à bớt bớt đi địt mẹ k có tiền ms buồn đi buồn vì 3 cái thằng vác buồi đi khắp thế gian lm lồn gì ns mãi vẫn thế đm ngu nó ít thôi cho t nhờ      1\n",
              "602                                                                                                                                                                                                        Hồ như dit tổ mày.rôi gia đinh may cug vay thoi.đeo co ti tinh nguoi.con cho      1\n",
              "2438                                                                                                                                                                                                                    2 th kia là 2 th sạo lồn , chỉ có tui là muốn bên bạn thật lòng      1\n",
              "1583                                                                                                                                                                                                            Bong Ma cái lồn mẹ mày thằng óc chó. Không cần là bác sĩ mới biết được.      1\n",
              "830                                                                                                                                                                                                                                Nguyen Cherry nao may co oc kg ? Ngu con noi Nhieu !      1\n",
              "2402                                                                                                                                                                                                                           Ngọc Ánh Bb như cc ck tôi rồi riêng cha này k nhường nhé      1\n",
              "769                                                                                                                                                                                                             Loại con gái như con này có chó nó mới ưng , xấu mà còn đóng vai ác nửa      1\n",
              "166                                                                                                          huyhy nguyễn thôi bạn ơi. Đừng có lên mạng xl về trải sự đời với thất bại và thành công với người khác.\\nKhông lại thành người hay nói đạo lý thì thường sống như l*n đấy.      1\n",
              "739                                                                                                                                      Nguyen Kenny .bắc kỳ chó. Bắc kỳ chúng mày đi theo CS. Nghèo nàn lạc hậu đói khổ chết mẹ. Đi ăn cướp miền nam rồi xạo lồn nói là đi giải phóng      1\n",
              "531                                                                                                                                                                                                       Yêu cc ty xạo lol \\nXấu ai theo \\nNghèo ai ngó \\nĐang iu cái lol ty bãi rác 🙂      1\n",
              "485                                                                                                                                                                                           Tou Prong Sép khá bảnh là bố của mày hả, cái địt cụ thằng óc chó này, cho mày ăn bả bây h      1\n",
              "1173                                                                                                                                      Nghiên Ngan ....mấy thằng bình luận ngu vl. Nó hỏi thích xoạc con gái ở đâu. Cứ trả lời lung tung. Thích xoạc ngã ba chứ chả lẽ xoạc sau lưng      1\n",
              "2148                                                      Đắng Cay Cuộc Đời bạn ns thế chịu bạn gặp trực tiếp ngta mà lm vc ngu cứ tỏ ra khôn lz. Ai ns nợ xấu ko vay đc bạn ko quen biết ai trong ngân hàng vào vay ai chả ko hỗ trợ. Quan hệ ko có, có những ng quan hệ tốt họ lo đc       1\n",
              "2699                                                                                                                                                                                               Trọng's Oppa's bố trẻ m còn dzai hơn th bố ông nội nhà m địt mẹ th quê , CÂM và CÚT       1\n",
              "988                                                                                                                   Hà S Ùng dm tau thấy hay mà.mày làm dk như vậy ko.mấy cái thể loại bà 62 tuổi lái phi công .hay thầy giáo hiếp hs.hotgirl lộ clip sao bọn mày ko bảo chất độc đi.      1\n",
              "1537                                                                                                                                                                                                                                  Dm~~ đồ ranh con dám hun ng thg của ta huhu 😭😂😂😂😍      1\n",
              "2255                                                                                                                                                                                                                                   con đĩ trung đứng píc pót :))) con meo đang nhảy      1\n",
              "975                                                                                                                                                                                                             Bong Ma cái lồn mẹ mày thằng óc chó. Không cần là bác sĩ mới biết được.      1\n",
              "632                                                                                                                                                 Đỡ hơn mấy ông chơi cứ văng tục, chửi người ta ngu này nọ, đằng này FL nháy đèn khịa đối thủ làm tăng độ hấp dẫn cho game thôi :))       1\n",
              "2540                                                                                                                                         Lủ trẻ trâu bây bớt sủa đi thằng Kẻ Gác Bình Minh nó nói đúng ròi .nhìn như lủ điên dm lủ trẩu con zô ẳng hả sủa như đúng rồi đúng não chó      1\n",
              "1119                                                                                                                                                                                                                        Tù đầu buồi. Nói như mày thanh niên việt nam đi tù gần hết.      1\n",
              "1755                                                                                                                                                                                            Bọn chó này nếu vào tay tôi thì tôi tử hình hết ko để một kẻ nào để hại Dân việt nam cả      1\n",
              "2309                                                                                                                                                                                                                      ngu vl.m cóc J t cóc Q cơ.xem m danh gì.ngu còn tỏ ra ngu hẳn      1"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-7c3eda97-ba9f-4cf8-a3a1-f42cc63b333e\" class=\"colab-df-container\">\n",
              "    <div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>text</th>\n",
              "      <th>label</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>1162</th>\n",
              "      <td>Phạm Thái Hoàng Long ồ loại m đ cãi dc là lại địt mẹ mày à, loạn luân vcl</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>161</th>\n",
              "      <td>Joseph Tuấn Anh T mà ăn hiếp m à thằng lồn</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1507</th>\n",
              "      <td>Anh Duy lại chó nữa🐶\\nCái dm chó cắn chó ăn cứt đi đừng có đi cắn người</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1920</th>\n",
              "      <td>Khanh Pham Van mày đi hỏi nó đi! Đm ngu lồn tỏ vẻ hiểu biết.</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1395</th>\n",
              "      <td>Má sử cho đẹp con đĩ đó nha má nghe nó ns mà tức thiệt chứ</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1211</th>\n",
              "      <td>Hiền Hén thông cảm cho con mụ điên đấy tý đi chị êii</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2106</th>\n",
              "      <td>Tui xin hỏi tụi con gái thử mày bị đánh trc mặt biết bao người v có biết nhục ko may cho mày là kiếm đc thằng ck bt nhịn mày gặp máy đứa khác đánh cho mày tung lôl</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2229</th>\n",
              "      <td>Dcm may thang lon chung may ngu vay. Bon cho cong an ko đánh nhung no goi nguòi khac toi đánh. Chúng may song ma éo biet gi chet het đi bon ngu</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2161</th>\n",
              "      <td>Hùng Sơn thằng đéo bò chữa hoang này.mày biết tao vùng nào không mà vào sủa  vậy? Dòng họ tao điều có 1 số người làm việc và theo 2 chế độ.tôn giáo cũng có thiên chúa nghen thằng đầu bò.nhưng khi hòa bình 2 phía vẫn nhìn nhau và giúp đỡ lẫn nhau</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>567</th>\n",
              "      <td>Hồng Thơm  hết chia sẻ rồi lại bình luận xàm lồn để lzi để chúng nó thấy rồi cười vào mặt cho à bớt bớt đi địt mẹ k có tiền ms buồn đi buồn vì 3 cái thằng vác buồi đi khắp thế gian lm lồn gì ns mãi vẫn thế đm ngu nó ít thôi cho t nhờ</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>602</th>\n",
              "      <td>Hồ như dit tổ mày.rôi gia đinh may cug vay thoi.đeo co ti tinh nguoi.con cho</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2438</th>\n",
              "      <td>2 th kia là 2 th sạo lồn , chỉ có tui là muốn bên bạn thật lòng</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1583</th>\n",
              "      <td>Bong Ma cái lồn mẹ mày thằng óc chó. Không cần là bác sĩ mới biết được.</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>830</th>\n",
              "      <td>Nguyen Cherry nao may co oc kg ? Ngu con noi Nhieu !</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2402</th>\n",
              "      <td>Ngọc Ánh Bb như cc ck tôi rồi riêng cha này k nhường nhé</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>769</th>\n",
              "      <td>Loại con gái như con này có chó nó mới ưng , xấu mà còn đóng vai ác nửa</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>166</th>\n",
              "      <td>huyhy nguyễn thôi bạn ơi. Đừng có lên mạng xl về trải sự đời với thất bại và thành công với người khác.\\nKhông lại thành người hay nói đạo lý thì thường sống như l*n đấy.</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>739</th>\n",
              "      <td>Nguyen Kenny .bắc kỳ chó. Bắc kỳ chúng mày đi theo CS. Nghèo nàn lạc hậu đói khổ chết mẹ. Đi ăn cướp miền nam rồi xạo lồn nói là đi giải phóng</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>531</th>\n",
              "      <td>Yêu cc ty xạo lol \\nXấu ai theo \\nNghèo ai ngó \\nĐang iu cái lol ty bãi rác 🙂</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>485</th>\n",
              "      <td>Tou Prong Sép khá bảnh là bố của mày hả, cái địt cụ thằng óc chó này, cho mày ăn bả bây h</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1173</th>\n",
              "      <td>Nghiên Ngan ....mấy thằng bình luận ngu vl. Nó hỏi thích xoạc con gái ở đâu. Cứ trả lời lung tung. Thích xoạc ngã ba chứ chả lẽ xoạc sau lưng</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2148</th>\n",
              "      <td>Đắng Cay Cuộc Đời bạn ns thế chịu bạn gặp trực tiếp ngta mà lm vc ngu cứ tỏ ra khôn lz. Ai ns nợ xấu ko vay đc bạn ko quen biết ai trong ngân hàng vào vay ai chả ko hỗ trợ. Quan hệ ko có, có những ng quan hệ tốt họ lo đc</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2699</th>\n",
              "      <td>Trọng's Oppa's bố trẻ m còn dzai hơn th bố ông nội nhà m địt mẹ th quê , CÂM và CÚT</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>988</th>\n",
              "      <td>Hà S Ùng dm tau thấy hay mà.mày làm dk như vậy ko.mấy cái thể loại bà 62 tuổi lái phi công .hay thầy giáo hiếp hs.hotgirl lộ clip sao bọn mày ko bảo chất độc đi.</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1537</th>\n",
              "      <td>Dm~~ đồ ranh con dám hun ng thg của ta huhu 😭😂😂😂😍</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2255</th>\n",
              "      <td>con đĩ trung đứng píc pót :))) con meo đang nhảy</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>975</th>\n",
              "      <td>Bong Ma cái lồn mẹ mày thằng óc chó. Không cần là bác sĩ mới biết được.</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>632</th>\n",
              "      <td>Đỡ hơn mấy ông chơi cứ văng tục, chửi người ta ngu này nọ, đằng này FL nháy đèn khịa đối thủ làm tăng độ hấp dẫn cho game thôi :))</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2540</th>\n",
              "      <td>Lủ trẻ trâu bây bớt sủa đi thằng Kẻ Gác Bình Minh nó nói đúng ròi .nhìn như lủ điên dm lủ trẩu con zô ẳng hả sủa như đúng rồi đúng não chó</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1119</th>\n",
              "      <td>Tù đầu buồi. Nói như mày thanh niên việt nam đi tù gần hết.</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1755</th>\n",
              "      <td>Bọn chó này nếu vào tay tôi thì tôi tử hình hết ko để một kẻ nào để hại Dân việt nam cả</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2309</th>\n",
              "      <td>ngu vl.m cóc J t cóc Q cơ.xem m danh gì.ngu còn tỏ ra ngu hẳn</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "    <div class=\"colab-df-buttons\">\n",
              "\n",
              "  <div class=\"colab-df-container\">\n",
              "    <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-7c3eda97-ba9f-4cf8-a3a1-f42cc63b333e')\"\n",
              "            title=\"Convert this dataframe to an interactive table.\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\" viewBox=\"0 -960 960 960\">\n",
              "    <path d=\"M120-120v-720h720v720H120Zm60-500h600v-160H180v160Zm220 220h160v-160H400v160Zm0 220h160v-160H400v160ZM180-400h160v-160H180v160Zm440 0h160v-160H620v160ZM180-180h160v-160H180v160Zm440 0h160v-160H620v160Z\"/>\n",
              "  </svg>\n",
              "    </button>\n",
              "\n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    .colab-df-buttons div {\n",
              "      margin-bottom: 4px;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "    <script>\n",
              "      const buttonEl =\n",
              "        document.querySelector('#df-7c3eda97-ba9f-4cf8-a3a1-f42cc63b333e button.colab-df-convert');\n",
              "      buttonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "      async function convertToInteractive(key) {\n",
              "        const element = document.querySelector('#df-7c3eda97-ba9f-4cf8-a3a1-f42cc63b333e');\n",
              "        const dataTable =\n",
              "          await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                    [key], {});\n",
              "        if (!dataTable) return;\n",
              "\n",
              "        const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "          '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "          + ' to learn more about interactive tables.';\n",
              "        element.innerHTML = '';\n",
              "        dataTable['output_type'] = 'display_data';\n",
              "        await google.colab.output.renderOutput(dataTable, element);\n",
              "        const docLink = document.createElement('div');\n",
              "        docLink.innerHTML = docLinkHtml;\n",
              "        element.appendChild(docLink);\n",
              "      }\n",
              "    </script>\n",
              "  </div>\n",
              "\n",
              "\n",
              "<div id=\"df-01b615c3-3184-4ff0-8445-418eb857dcff\">\n",
              "  <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-01b615c3-3184-4ff0-8445-418eb857dcff')\"\n",
              "            title=\"Suggest charts\"\n",
              "            style=\"display:none;\">\n",
              "\n",
              "<svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "     width=\"24px\">\n",
              "    <g>\n",
              "        <path d=\"M19 3H5c-1.1 0-2 .9-2 2v14c0 1.1.9 2 2 2h14c1.1 0 2-.9 2-2V5c0-1.1-.9-2-2-2zM9 17H7v-7h2v7zm4 0h-2V7h2v10zm4 0h-2v-4h2v4z\"/>\n",
              "    </g>\n",
              "</svg>\n",
              "  </button>\n",
              "\n",
              "<style>\n",
              "  .colab-df-quickchart {\n",
              "      --bg-color: #E8F0FE;\n",
              "      --fill-color: #1967D2;\n",
              "      --hover-bg-color: #E2EBFA;\n",
              "      --hover-fill-color: #174EA6;\n",
              "      --disabled-fill-color: #AAA;\n",
              "      --disabled-bg-color: #DDD;\n",
              "  }\n",
              "\n",
              "  [theme=dark] .colab-df-quickchart {\n",
              "      --bg-color: #3B4455;\n",
              "      --fill-color: #D2E3FC;\n",
              "      --hover-bg-color: #434B5C;\n",
              "      --hover-fill-color: #FFFFFF;\n",
              "      --disabled-bg-color: #3B4455;\n",
              "      --disabled-fill-color: #666;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart {\n",
              "    background-color: var(--bg-color);\n",
              "    border: none;\n",
              "    border-radius: 50%;\n",
              "    cursor: pointer;\n",
              "    display: none;\n",
              "    fill: var(--fill-color);\n",
              "    height: 32px;\n",
              "    padding: 0;\n",
              "    width: 32px;\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart:hover {\n",
              "    background-color: var(--hover-bg-color);\n",
              "    box-shadow: 0 1px 2px rgba(60, 64, 67, 0.3), 0 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "    fill: var(--button-hover-fill-color);\n",
              "  }\n",
              "\n",
              "  .colab-df-quickchart-complete:disabled,\n",
              "  .colab-df-quickchart-complete:disabled:hover {\n",
              "    background-color: var(--disabled-bg-color);\n",
              "    fill: var(--disabled-fill-color);\n",
              "    box-shadow: none;\n",
              "  }\n",
              "\n",
              "  .colab-df-spinner {\n",
              "    border: 2px solid var(--fill-color);\n",
              "    border-color: transparent;\n",
              "    border-bottom-color: var(--fill-color);\n",
              "    animation:\n",
              "      spin 1s steps(1) infinite;\n",
              "  }\n",
              "\n",
              "  @keyframes spin {\n",
              "    0% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "      border-left-color: var(--fill-color);\n",
              "    }\n",
              "    20% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    30% {\n",
              "      border-color: transparent;\n",
              "      border-left-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    40% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-top-color: var(--fill-color);\n",
              "    }\n",
              "    60% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "    }\n",
              "    80% {\n",
              "      border-color: transparent;\n",
              "      border-right-color: var(--fill-color);\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "    90% {\n",
              "      border-color: transparent;\n",
              "      border-bottom-color: var(--fill-color);\n",
              "    }\n",
              "  }\n",
              "</style>\n",
              "\n",
              "  <script>\n",
              "    async function quickchart(key) {\n",
              "      const quickchartButtonEl =\n",
              "        document.querySelector('#' + key + ' button');\n",
              "      quickchartButtonEl.disabled = true;  // To prevent multiple clicks.\n",
              "      quickchartButtonEl.classList.add('colab-df-spinner');\n",
              "      try {\n",
              "        const charts = await google.colab.kernel.invokeFunction(\n",
              "            'suggestCharts', [key], {});\n",
              "      } catch (error) {\n",
              "        console.error('Error during call to suggestCharts:', error);\n",
              "      }\n",
              "      quickchartButtonEl.classList.remove('colab-df-spinner');\n",
              "      quickchartButtonEl.classList.add('colab-df-quickchart-complete');\n",
              "    }\n",
              "    (() => {\n",
              "      let quickchartButtonEl =\n",
              "        document.querySelector('#df-01b615c3-3184-4ff0-8445-418eb857dcff button');\n",
              "      quickchartButtonEl.style.display =\n",
              "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "    })();\n",
              "  </script>\n",
              "</div>\n",
              "\n",
              "    </div>\n",
              "  </div>\n"
            ],
            "application/vnd.google.colaboratory.intrinsic+json": {
              "type": "dataframe",
              "summary": "{\n  \"name\": \"test\",\n  \"rows\": 32,\n  \"fields\": [\n    {\n      \"column\": \"text\",\n      \"properties\": {\n        \"dtype\": \"string\",\n        \"num_unique_values\": 31,\n        \"samples\": [\n          \"L\\u1ee7 tr\\u1ebb tr\\u00e2u b\\u00e2y b\\u1edbt s\\u1ee7a \\u0111i th\\u1eb1ng K\\u1ebb G\\u00e1c B\\u00ecnh Minh n\\u00f3 n\\u00f3i \\u0111\\u00fang r\\u00f2i .nh\\u00ecn nh\\u01b0 l\\u1ee7 \\u0111i\\u00ean dm l\\u1ee7 tr\\u1ea9u con z\\u00f4 \\u1eb3ng h\\u1ea3 s\\u1ee7a nh\\u01b0 \\u0111\\u00fang r\\u1ed3i \\u0111\\u00fang n\\u00e3o ch\\u00f3\",\n          \"Lo\\u1ea1i con g\\u00e1i nh\\u01b0 con n\\u00e0y c\\u00f3 ch\\u00f3 n\\u00f3 m\\u1edbi \\u01b0ng , x\\u1ea5u m\\u00e0 c\\u00f2n \\u0111\\u00f3ng vai \\u00e1c n\\u1eeda\",\n          \"H\\u00e0 S \\u00d9ng dm tau th\\u1ea5y hay m\\u00e0.m\\u00e0y l\\u00e0m dk nh\\u01b0 v\\u1eady ko.m\\u1ea5y c\\u00e1i th\\u1ec3 lo\\u1ea1i b\\u00e0 62 tu\\u1ed5i l\\u00e1i phi c\\u00f4ng .hay th\\u1ea7y gi\\u00e1o hi\\u1ebfp hs.hotgirl l\\u1ed9 clip sao b\\u1ecdn m\\u00e0y ko b\\u1ea3o ch\\u1ea5t \\u0111\\u1ed9c \\u0111i.\"\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    },\n    {\n      \"column\": \"label\",\n      \"properties\": {\n        \"dtype\": \"number\",\n        \"std\": 0,\n        \"min\": 1,\n        \"max\": 1,\n        \"num_unique_values\": 1,\n        \"samples\": [\n          1\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    }\n  ]\n}"
            }
          },
          "metadata": {},
          "execution_count": 42
        }
      ]
    }
  ]
}