Document-level relation extraction (DocRE) task aims to identify relations between entities in a document. In DocRE, an entity may appear in multiple sentences of a document in the form of mentions. In addition, relation inference requires the use of evidence sentences that can provide key clues to entity pairs. These make DocRE more challenging than sentencelevel relation extraction. Existing work does not fully distinguish the contribution of different mentions to entity representation and the importance of mentions in evidence sentences. To address these issues, we observe that entity types can provide consistent semantic constraints for entities of the same type and implicitly preclude impossible relations between entities, which may help the model better understand both intra- and inter-entity mentions. Therefore, we propose a novel model ET-MIER, which for the first time leverages **E**ntity **T**ypes to guide key **M**ention **I**dentification and **E**vidence **R**etrieval. In this way, entity types not only help learn better entity representation but also enhance evidence retrieval, both of which are crucial for DocRE. We conduct experiments on widely-adopted datasets and show that our model achieves state-of-the-art performance. Our code is available at: https://github.com/NEU-IDKE/ET-MIER

Document-level Relation Extraction (DocRE) aims to extract relations between entity pairs in a document and poses many challenges as it involves multiple mentions of entities and cross-sentence inference. However, several aspects that are important for DocRE have not been considered and explored. Existing work ignores bidirectional mention interaction when generating relational features for entity pairs. Also, sophisticated neural networks are typically designed for cross-sentence evidence extraction to further enhance DocRE. More interestingly, we reveal a noteworthy finding: If a model has predicted a relation between an entity and other entities, this relation information may help infer and predict more relations between the entity’s adjacent entities and these other entities. Nonetheless, none of existing methods leverage secondary reasoning to exploit results of relation prediction. To this end, we propose a novel Secondary Reasoning Framework (SRF) for DocRE. In SRF, we initially propose a DocRE model that incorporates bidirectional mention fusion and a simple yet effective evidence extraction module (incurring only an additional learnable parameter overhead) for relation prediction. Further, for the first time, we elaborately design and propose a novel secondary reasoning method to discover more relations by exploring the results of the first relation prediction. Extensive experiments show that SRF achieves SOTA performance and our secondary reasoning method is both effective and general when integrated into existing models.