Unleashing the power of context: Contextual association network with cross-task attention for joint relational extraction

Da Luo, Yanglei Gan, Rui Hou, Qiao Liu, Tingting Dai, Yuxiang Cai, Xiaojun Shi

Abstract: Highlights•We propose a novel approach for joint relational triple extraction that leverages contextual information.•We employ a cross-task attention mechanism to addresses the issue of rote memorization and implicit relation.•The model can handle multiple relation types and overlapping entity mentions within a sentence.•Experiments on four widely-used benchmarks verifies the effectiveness of our model.