crack521 / word-cluster-embedding

Code for "Smaller Text Classifiers with Discriminative Cluster Embeddings" (NAACL 2018)

Word Cluster Embeddings

Code for "Smaller Text Classifiers with Discriminative Cluster Embeddings" (NAACL 2018)

Citation

If you use our code, please cite:

@inproceedings{chen2018smaller,
  author = {Mingda Chen and Kevin Gimpel},
  booktitle = {North American Association for Computational Linguistics (NAACL)},
  title = {Smaller Text Classifiers with Discriminative Cluster Embeddings},
  year = {2018}
}

Dependencies

• Python 3.5
• TensorFlow 1.3
• NLTK (for tokenizing IMDB dataset)

Prepare Data

You can download AG News, DBpedia, Yelp Review Full, Yelp Review Polarity from here and IMDB data from here. Then run the corresponding data processing code to generate data file.

Note that in this code Yelp Review Full and Yelp Review Polarity were renamed to yelp-1 and yelp-2 respectively. If you want to use other names, please modify the code accordingly.

Training

Use run_se_ce_cae.py for training standard embeddings, cluster embeddings or cluster adjustment embeddings. Use run_me.py for training mixture embeddings.