/sequential_gnn_for_script_event_prediction

Primary LanguagePython

SeqGNN

Introduction

The codes here include PyTorch implementations of the baseline and our SeqGNN model. Our code is based on SGNN. Code for EventComp model and how to extract the narrative event chains from raw NYT news corpus can be found here.

Environmental dependence

  • Python 3.7.4
  • PyTorch 1.3.3
  • Red Hat 4.8.5-28
  • GPU (TITAN V)

How to run the code?

First, you need to download the data and put it in the data folder. Data includes deepwalk_128_unweighted_with_args.txt.

Second, you can config the parameters of the model by config.py and run main.py to train the model.

python main.py # train the seqgnn model

Third, you can run chain.py to train the baseline of SeqGNN-GRUFusion.

python chain.py -l 4 -m train # train the 4th location
python chain.py -l 4 -m test # test the 4th location

Fourth, you can run evaluate.py to evaluate the accuracy of seqgnn model.

python evaluate.py