BTDM: A Bi-directional Translating Decoding Model Based On Relational Embedding For Relational Triple Extraction

This repository contains the source code and dataset for the paper: BTDM: A Bi-directional Translating Decoding Model Based On Relational Embedding For Relational Triple Extraction.

Overview

Requirements

The main requirements are:

• Keras==2.3.1
• numpy==1.23.5
• tensorflow==2.3.1
• torch==1.12.1
• transformers==4.25.1
• tqdm

Datasets

• NYT* and WebNLG*(following CasRel)
• NYT(following CopyRE)
• WebNLG(following ETL-span)

Usage

1. Get pre-trained BERT model for PyTorch

Download BERT-Base-Cased which contains pytroch_model.bin, vocab.txt and config.json. Put these under ./pretrained.

2. Build Data

Put our preprocessed datasets under ./datasets.

3. Train

Specify the running mode and dataset at the command line

python run.py ---train=train --dataset=NYT

4. Evaluate

Specify the running mode and dataset at the command line

python run.py ---train=test --dataset=NYT