A Text Matching-based Knowledge Fusion Method For Power Metering Domain

In the process of constructing the knowledge graph in the domain of electric power metering, the knowledge often comes from several different data sources, and because of the differences in structure and content among the data sources, there are also problems of multiple meanings of words and unclear referents in the integration of the acquired knowledge. Therefore, knowledge fusion processing is needed to eliminate the ambiguity of entities and attributes in terms of denotation. This paper proposes a knowledge fusion method based on text matching technology. The method includes three core steps. Firstly it use a crawler system to obtain corpus data. Secondly it use the BERT-based knowledge extraction model to extract knowledge and build the entity information knowledge base. Finally the entities are matched at the semantic level to complete the knowledge fusion process. The experimental results prove that the knowledge fusion effect of the method in this paper is good and the entity matching rate is high.