基于多数据源的知识图谱构建方法研究

针对多数据源的融合应用,构建了基于多数据源的知识图谱.首先,对不同领域内的数据源构建相应本体库,并将不同本体库通过数据融合映射到全局本体库;然后,利用实体对齐和实体链接方法进行知识获取和融合;最后,搭建知识图谱应用平台,提供查询和统计等操作.在实体对齐方面,利用传统的基于相似性传播实体对齐方法,获得良好的实体对齐效果;在实体链接方面,提出了基于约束嵌入转换的预测推理方法,实验结果表明,在预测准确率上取得较好的结果.

Knowledge graph construction method based on multiple data sources

To improve the application of multi-source data fusion, this study put forward a knowledge graph-based data fusion construction model. This model firstly constructed corresponding domain ontology for each special field, and then consolidated all domain ontology into a global ontology. After that, it retrieved and fused knowledge from the global ontology by entity alignment and linking methods. At last it built an application platform of knowledge graph with friendly interfaces to execute query and statistics, etc. Besides that, this model improved the result of entity aligning by adopting traditional similarity detection approach. And experiment results also demonstrated its good prediction accuracy by proposing a constraint based embedding model in entity linking process.