This repo contains the code for a state-of-the-art discourse relation classifier, trained and tested on PDTB V2.

Here We implemented the Tree-LSTM and Tree-GRU models and enhanced them with pos-tag information. You can find more details in our paper at IJCNLP 2017: Tag-Enhanced Tree-Structured Neural Networks for Implicit Discourse Relation Classification.

We trained and and tested our models on PDTB V2 dataset. You should download them and convert the PDTB into a "pipe" delimited file using the built-in tool in the dataset.

To run the program, you can use the following command for preprocessing the data, prepare the formatted dataset and train the models:

python main.py [--preprocess] [--prepare] [--train]

For detailed usage, run python main.py -h to see the list of options.

MIT license