Instructions

---

For training:

Step-1: You need to collect the tweets for tweet ids mentioned in the data file, CRED_without_tweets.xlsx, and put them in the corresponding columns (i.e. 'cleaned_tweet_text_x' and 'cleaned_tweet_text_y'). Following Twitter's policy, we couldn't share the individual tweets. However, this process is simple and straight forward (Reference: https://stackoverflow.com/questions/28384588/twitter-api-get-tweets-with-specific-id). You need to clean the tweets (remove URLs) using the tweet-preprocessor library. Save the final file as CRED_with_tweets.xlsx.
Step-2: Install the packages mentioned in the requirements.txt file.
Step-3: Execute the notebook GGEA_Training_Script.ipynb following the instructions present in it. This notebook will create the relevant model artefacts.

For scoring pre-trained GGEA

Step-1: Install the packages mentioned in the requirements.txt file.
Step-2: Execute the notebook Scoring_GGEA.ipynb following the instructions present in it. This notebook will automatically download the model artefacts from HugingFace. It has been developed & tested using Google Colab.

@INPROCEEDINGS{ghosh-wtal-www2024,
  author={Ghosh, Sohom and Chen, Chung-Chi and Naskar, Sudip Kumar},
  booktitle={TheWebConf 2024}, 
  title={Generator-Guided Crowd Reaction Assessment}, 
  month={05},
  year={2024},
  volume={},
  number={},
  pages={},
  url={https://dl.acm.org/doi/10.1145/3589335.3651512}
}