NER_Drug_Prescriptions

A solution for the TAC 2017 - Adverse Drug Reaction Extraction from Drug Labels.

Architecture

Model implementation present in main_tf.py

1. Glove Embeddings for words
2. Character Embeddings
3. 1d convolution and max pooling on Character Embeddings
4. Bi-LSTM
5. CRF

Data

The data files are downloaded from here

Download glove embeddings from here. Extract it inside the data folder.

Scripts

1. annotate_data.py - Preprocess and tag data.
2. build_vocab.py - Build the vocabulary
3. build_glove.py - Retrieve glove embeddings for each word.