基于子图相交的社交账号与知识图谱实体对齐

Subgraph Intersection Based Alignment between Social Media Account and Knowledge Graph Entity

社交媒体与知识图谱的数据各具特点,相互之间的数据互通具有较强的现实意义,而社交账号与知识图谱实体的对齐是数据互通的前提。针对社交媒体与知识图谱的特点,提出了一种基于子图相交的对齐方法,旨在给定社交账号的情况下,根据社交账号的相关信息在知识图谱中找到正确的对应条目。该方法在候选实体生成阶段对比实验了不同的生成策略。在目标实体选择阶段提出一种基于子图相交的算法,利用社交账号的社交关系在知识图谱中映射成子图。子图相交算法通过考察子图中候选实体周围顶点的"稠密"程度,确定社交账号所对应的目标实体。由于该领域尚无公开可用的测试数据集,构造了一个基于Twitter与Wikidata的对齐数据集,使用该数据集对该方法进行评估,对比测试了标题匹配算法和AGDISTIS算法,子图相交算法能够达到更好的效果。

The data of social media and knowledge graph have their own characteristics,and the data exchange between them has strong practical significance. The alignment of social accounts and knowledge graph entities is the premise of data exchange. Focused on the characteristics of social media and knowledge graph,an alignment method based on subgraph intersection is proposed to find the correct corresponding entries in knowledge graph under given social media accounts. In the phase of candidate entity stage,different generation strategies are compared and experimented. A subgraph intersection algorithm in the target entity selection stage is presented,which creates subgraphs by using the social relations of social media accounts within knowledge graph. By investigating the "density" of vertices around candidate entities in the subgraph,the target entities corresponding to social media accounts are selected. There is no publicly available data set for testing and evaluation in this field,so an aligned data set based on Twitter and Wikidata is constructed to evaluate the proposed method,and the algorithm based on title comparison and AGDISTIS algorithm are tested. Subgraph intersection method can achieve better results.