A solution for the TAC 2017 - Adverse Drug Reaction Extraction from Drug Labels.
Model implementation present in main_tf.py
- Glove Embeddings for words
- Character Embeddings
- 1d convolution and max pooling on Character Embeddings
- Bi-LSTM
- CRF
The data files are downloaded from here
Download glove embeddings from here. Extract it inside the data folder.
- annotate_data.py - Preprocess and tag data.
- build_vocab.py - Build the vocabulary
- build_glove.py - Retrieve glove embeddings for each word.