Web application for exploring the vector additive properties of word2vec
Gensim, a Python library used to do semantic analyis of text, is useed with Google's News Vectors Dataset to provide the backend of this app.
The frontend is build using VueJS.
Users are given a target word that they must build using a word equation. For example:
king - man + woman = queen
This is possible because of the vector additive properties of word embeddings. Though there are a few caveats to describing word embeddings as living in a 'vector space', it is still a good descriptor.
The backend is a Flask application. You can run it with Docker with:
cd backend
docker-compose upNote that you must have a model present in the models directory for the app to run when using Docker. The Google Cloud SDK version will use a hosted version of the model, so you don't have to worry if running that way.
The model file itself that is running in production can be found here
The frontend is a Vue.js single page application.
To run locally,
cd frontend
yarn install
yarn run production # you can use yarn run local if running against local backend