Web实体关系抽取算法研究

Research on the Web Entity Relationships Extraction Algorithm

文章提出一种基于多模态注意力网络(Span-based Multi-Modal Attention Network,SMAN)的实体关系联合抽取算法,该算法通过完形填空机制建模文本上下文信息,并利用模态增强注意力模块(Modal Enhancement Attention Module,MEA)捕捉跨模态数据间的细粒度交互特征。实验表明,SMAN算法在Web数据上能显著提高实体识别和关系抽取的性能,为构建高质量的Web知识图谱提供了有效方法。

This paper proposes a joint entity relation extraction algorithm based on multimodal attention Network(SMAN).The algorithm utilizes cloze mechanism to model text context information and employs Modal Enhanced Attention module(MEA)to capture fine-grained interaction features between cross-modal data.Experiments demonstrate that the SMAN algorithm can significantly enhance the performance of entity recognition and relationship extraction on Web data,offering an effective approach for constructing high-quality Web knowledge graph.