Document-level relation extraction based on sememe knowledge-enhanced abstract meaning representation and reasoning

Abstract Document-level relation extraction is a challenging task in information extraction, as it involves identifying semantic relations between entities that are dispersed throughout document. Existing graph-based approaches often rely on simplistic methods to construct text graphs, which do not provide enough lexical and accurately predict the entity pairs. In this paper, we introduce document-level method called SKAMRR ( S ememe K nowledge-enhanced A bstract M eaning R epresentation easoning). First, generate abstract meaning representation graphs using rules acquire nodes’ features through sufficient propagation. Next, inference for pairs utilize graph neural networks obtain their representations classification. Additionally, propose global adaptive loss address issue of long-tailed data. We conduct extensive experiments four datasets DocRE, CDR, GDA, HacRED. Our model achieves competitive results its performance outperforms previous state-of-the-art datasets.