Home

Awesome

Classifier Decomposition

Codes for our paper "Enhancing Continual Relation Extraction via Classifier Decomposition (Findings of ACL2023). In this paper:

cdec

Environment

conda create -n cdec python=3.8
conda activate cdec
pip install -r requirements.txt

Dataset

We use two datasets in our experiments, FewRel and TACRED:

We construct 5 CRE task sequences in both FewRel and TACRED the same as RP-CRE and CRL.

Run

bash ./bash/[dataset]/cdec.sh
    - dataset: the dataset name, e.g.,:
        - fewrel/tacred

Note

This code is based on ACA (https://github.com/Wangpeiyi9979/ACA).

Citation

@inproceedings{Xia:2023CDec,
  author       = {Heming Xia and
                  Peiyi Wang and
                  Tianyu Liu and
                  Binghuai Lin and
                  Yunbo Cao and
                  Zhifang Sui},
  editor       = {Anna Rogers and
                  Jordan L. Boyd{-}Graber and
                  Naoaki Okazaki},
  title        = {Enhancing Continual Relation Extraction via Classifier Decomposition},
  booktitle    = {Findings of the Association for Computational Linguistics: {ACL} 2023,
                  Toronto, Canada, July 9-14, 2023},
  pages        = {10053--10062},
  publisher    = {Association for Computational Linguistics},
  year         = {2023},
  url          = {https://doi.org/10.18653/v1/2023.findings-acl.638},
  doi          = {10.18653/V1/2023.FINDINGS-ACL.638},
  timestamp    = {Thu, 10 Aug 2023 12:35:45 +0200},
  biburl       = {https://dblp.org/rec/conf/acl/XiaW0LCS23.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}