Agent-Like Model for the Fusion of Attention Features for the Extraction of Joint-Entity Relations

Relation extraction involves identifying related entity pairs within sentences and matching them with corresponding relation types. This paper introduces an agent-like model that fuses attention features to facilitate relation extraction. Based on a cascade binary tagging framework, the model uses an agent-like module enabling the efficient extraction of relations and implicit semantic information from training data. In experiments, the proposed model improved efficiency in extracting relational triples.