English

Reasoning with Latent Structure Refinement for Document-Level Relation Extraction

Computation and Language 2020-07-29 v3

Abstract

Document-level relation extraction requires integrating information within and across multiple sentences of a document and capturing complex interactions between inter-sentence entities. However, effective aggregation of relevant information in the document remains a challenging research question. Existing approaches construct static document-level graphs based on syntactic trees, co-references or heuristics from the unstructured text to model the dependencies. Unlike previous methods that may not be able to capture rich non-local interactions for inference, we propose a novel model that empowers the relational reasoning across sentences by automatically inducing the latent document-level graph. We further develop a refinement strategy, which enables the model to incrementally aggregate relevant information for multi-hop reasoning. Specifically, our model achieves an F1 score of 59.05 on a large-scale document-level dataset (DocRED), significantly improving over the previous results, and also yields new state-of-the-art results on the CDR and GDA dataset. Furthermore, extensive analyses show that the model is able to discover more accurate inter-sentence relations.

Keywords

Cite

@article{arxiv.2005.06312,
  title  = {Reasoning with Latent Structure Refinement for Document-Level Relation Extraction},
  author = {Guoshun Nan and Zhijiang Guo and Ivan Sekulić and Wei Lu},
  journal= {arXiv preprint arXiv:2005.06312},
  year   = {2020}
}

Comments

Appeared in the proceedings of ACL 2020 (Long paper)

R2 v1 2026-06-23T15:30:54.590Z