Mention-centered Graph Neural Network for Document-level Relation Extraction
This work addresses the challenge of extracting relations between entities across entire documents for natural language processing applications, representing an incremental improvement over existing methods.
The paper tackles document-level relation extraction by building cross-sentence dependencies through compositional relations between mentions, using mention convolution and an improved ranking loss to address generalization issues from incomplete annotation. Experiments show this approach enables extraction of meaningful higher-level relations, with specific performance gains reported in benchmarks.
Document-level relation extraction aims to discover relations between entities across a whole document. How to build the dependency of entities from different sentences in a document remains to be a great challenge. Current approaches either leverage syntactic trees to construct document-level graphs or aggregate inference information from different sentences. In this paper, we build cross-sentence dependencies by inferring compositional relations between inter-sentence mentions. Adopting aggressive linking strategy, intermediate relations are reasoned on the document-level graphs by mention convolution. We further notice the generalization problem of NA instances, which is caused by incomplete annotation and worsened by fully-connected mention pairs. An improved ranking loss is proposed to attend this problem. Experiments show the connections between different mentions are crucial to document-level relation extraction, which enables the model to extract more meaningful higher-level compositional relations.