DWIE: An entity-centric dataset for multi-task document-level information extraction

作者：

Highlights：

• New DWIE dataset for joint NER, relation extraction, coreference, entity linking.

• Data-driven bottom-up, entity-centric annotation reflecting the content of the corpus.

• Graph-based model starting from entity spans with propagation along relation-edges.

• Our joint multi-task neural network models outperform single task models.

• Attention-driven graph propagation outperforms annotation-driven counterparts.

摘要

•New DWIE dataset for joint NER, relation extraction, coreference, entity linking.•Data-driven bottom-up, entity-centric annotation reflecting the content of the corpus.•Graph-based model starting from entity spans with propagation along relation-edges.•Our joint multi-task neural network models outperform single task models.•Attention-driven graph propagation outperforms annotation-driven counterparts.

论文关键词：Named entity recognition,Entity linking,Relation extraction,Coreference resolution,Joint models,Graph Neural Networks

论文评审过程：Received 17 August 2020, Revised 27 January 2021, Accepted 27 February 2021, Available online 20 March 2021, Version of Record 20 March 2021.

论文官网地址：https://doi.org/10.1016/j.ipm.2021.102563