Relation Extraction with The Wall Street Journal
- Gensim
- Keras
- sklearn
- Transfer documents into feature vectors by 3-billion words Google News pre-trained word2vec model
- Train a Bidirectional LSTM following DNN model to predict Node Role/Label
- Use the output of Bidirectional LSTM layer as input to train a DNN model for edge label prediction