Unofficial implementation of the paper Bag of Tricks for Efficient Text Classification by Joulin et al.

FastText requires Python 3 with Keras installed.

Obtain the Yelp Dataset from here and place yelp_academic_dataset_review.json in the base directory.

Train the model using the following command:

./train.py

It generates data.csv which represents the model's embedding space of the validation set. It is obtained by removing the last layer of the model and using t-SNE for the dimensionality reduction.

index.html implements a D3 visualisation to view the embedding space. You need to run a local web server because browsers don't allow file accesses:

python -m http.server 8000

Now point your browser to: localhost:8000.

FastText is licensed under the terms of the Apache v2.0 license.

• Ihor Kroosh
• Tim Nieradzik