Title:KnowRA: Knowledge Retrieval Augmented Method for Document-level Relation Extraction with Comprehensive Reasoning Abilities

Abstract:Document-level relation extraction (Doc-RE) aims to extract relations between entities across multiple sentences. Therefore, Doc-RE requires more comprehensive reasoning abilities like humans, involving complex cross-sentence interactions between entities, contexts, and external general knowledge, compared to the sentence-level RE. However, most existing Doc-RE methods focus on optimizing single reasoning ability, but lack the ability to utilize external knowledge for comprehensive reasoning on long documents. To solve these problems, a knowledge retrieval augmented method, named KnowRA, was proposed with comprehensive reasoning to autonomously determine whether to accept external knowledge to assist DocRE. Firstly, we constructed a document graph for semantic encoding and integrated the co-reference resolution model into KnowRA to augment the co-reference reasoning ability. Then, we further expanded the document graph into a document knowledge graph by retrieving the external knowledge base and introduced the axis attention mechanism into KnowRA to improve its common-sense and logical reasoning abilities, respectively. Finally, a knowledge filtering method was presented in the common-sense and co-reference reasoning module to filter out irrelevant knowledge. Extensive experiments conducted on two datasets verified the effectiveness of our method compared to the state-of-the-art baselines. Our code is available at this https URL.