This project provides Tensorflow based implementation of LSTM-CNN model for n-ary relation extraction for the paper, Combining Long Short Term Memory and Convolutional Neural Network for Cross-Sentence n-ary Relation Extraction.
- Python 2.7
- Tensorflow >= 1.6
This paper uses dataset developed by Peng et al., 2016.
The original dataset is preprocessed for the LSTM-CNN model and is provided in data.zip. Unzip data.zip to obtain 'data' folder consisting preprocessed data for different relations.
The model uses Glove pre-trained embeddings. Download glove.6B.300d.txt into 'glove' floder.
Five-fold cross-validation is conducted. The model is run on different folds using the following and the average is taken across five folds:
- python train_lstm_cnn.py data/drug_var_cross_sents/fold_0/ ./training_config.json