Skip to content

Latest commit

 

History

History
28 lines (14 loc) · 825 Bytes

README.md

File metadata and controls

28 lines (14 loc) · 825 Bytes

This repository contains the source code of the paper "A Hierarchical Entity Graph Convolutional Network for Relation Extraction across Documents" published in RANLP 2021.

Datasets

The dataset used for experiments in the paper can be downloaded from the following link:

https://drive.google.com/drive/folders/1nBima4wLXSRHtMfcMdcQgu3AFlMe9AWP?usp=sharing

Requirements

  1. python3.6
  2. pytorch 1.7
  3. CUDA 8.0

How to run

python3.6 models.py source_dir embedding_file target_dir model_id train

embedding_file = path for glove.6B.300d.txt

python3.6 models.py source_dir embedding_file target_dir model_id test threshold

Above threshold value will be one of outputs of the training process

Use model_id as 1 for CNN, 2 for BiLSTM, 3 for BiLSTM_CNN, 4 for LinkPath, and 5 for our HEGCN model.