Skip to content

EMNLP2020 findings paper: Minimize Exposure Bias of Seq2Seq Models in Joint Entity and Relation Extraction

Notifications You must be signed in to change notification settings

WindChimeRan/OpenJERE

Repository files navigation

OpenJREE: Joint Relations and Entities Extraction

This is for EMNLP2020 findings paper: Minimize Exposure Bias of Seq2Seq Models in Joint Entity and Relation Extraction

Requirement

  • python 3.7/3.8
  • pytorch 1.6
pip install -r requirements.txt

Models

Install

pip install -e .

Run

Download the DuIE dataset from official website

Unzip *.json into ./raw_data/chinese/

For NYT, see raw_data/nyt/README.md

pip install -r requirements.txt
cd raw_data
unzip ../raw_data_joint.zip

Then run data_split for both datasets:

python data_split.py
bash train_all.sh

seperate steps:

python main.py --mode preprocessing --exp chinese_seq2umt_ops
python main.py --mode train --exp chinese_seq2umt_ops
python main.py --mode evaluation --exp chinese_seq2umt_ops
python main.py --mode preprocessing --exp nyt_seq2umt_ops
python main.py --mode train --exp nyt_seq2umt_ops
python main.py --mode evaluation --exp nyt_seq2umt_ops
python main.py --mode preprocessing --exp nyt_wdec
python main.py --mode train --exp nyt_wdec
python main.py --mode evaluation --exp nyt_wdec

About

EMNLP2020 findings paper: Minimize Exposure Bias of Seq2Seq Models in Joint Entity and Relation Extraction

Topics

Resources

Stars

Watchers

Forks

Packages

No packages published