swaggyzhang / continualre

ContinualRE

Codes and datasets for our paper "Continual Relation Learning via Episodic Memory Activation and Reconsolidation"

If you use the code, please cite the following paper:

 @inproceedings{han2018neural,
   title={Continual Relation Learning via Episodic Memory Activation and Reconsolidation},
   author={Han, Xu and Dai, Yi and Gao, Tianyu and Lin, Yankai and Liu, Zhiyuan and Li, Peng and Sun, Maosong and Zhou, Jie},
   booktitle={Proceedings of ACL},
   year={2020}
 }

Requirements

The model is implemented using PyTorch. The versions of packages used are shown below.

• numpy==1.18.0

• scikit-learn==0.22.1

• scipy==1.4.1

• torch==1.3.0

• tqdm==4.41.1

Baselines

The main experimental settings come from the project [Lifelong Relation Detection](https://github.com/hongwang600/ Lifelong_Relation_Detection).

We adapt some typical lifelong learning methods for continual relation learning, including EMR, AGEM and EWC. The code of these models can be found in the folder "./baseline/".

Datasets

We provide all the datasets and word embeddings used in our experiments.

• [Download (datasets)]

• [Download (word embeddings)]

Run the experiments

(0) To run the experiments, unpack the datasets and word embeddings first

unzip data.zip -d data/
unzip glove.zip -d glove/

(1) For FewRel

cp -r data/ fewrel/
cp -r glove/ fewrel/
cd fewrel
python run_multi_proto.py

(2) For SimpleQuestions

cp -r data/ simque/
cp -r glove/ simque/
cd simque
python run_multi_proto.py

(3) For TACRED

cp -r data/ tacred/
cp -r glove/ tacred/
cd tacred
python run_multi_proto.py

(4) For some special settings

All the config files can be found in "./fewrel/config/", "./tacred/config/", and "./simque/config/". By changing the config file name in the code "run_multi_proto.py", we can run experiments with different settings. In "./fewrel/config/", "./tacred/config/", and "./simque/config/", we also provide code to generate customized settings.