•  
  •  
 

Scientific Information Research

Keywords

domain knowledge graph; named entity recognition; entity relationship extraction; machine learning; electricity domain text; intelligent decision support

Abstract

[Purpose/significance] This research constructed a domain knowledge graph and its scenario-oriented application framework for decision support at four levels: the data foundation layer, the key technology layer, the domain knowledge graph construction layer, and the scenario-oriented application layer. This framework aims to provide systematic support for knowledge discovery.

[Method/process]Based on the construction of a domain knowledge graph and its scenario-oriented application framework for decision support, this research focuses on the improvement of models and performance evaluation for fine-grained entity and relationship extraction at the discourse level within texts. The optimal model is selected to construct a domain knowledge graph. Taking the field of transformer equipment failure as an example, empirical studies are conducted to realize scenario-oriented applications aimed at decision support.

[Result/conclusion]The integrated pipeline knowledge extraction method, which combines the improved BERT-BiLSTM CRF model and the PURE-RE model, has demonstrated superior comprehensive performance. The knowledge graph constructed for the domain of transformer faults has been verified against the expert knowledge in the "Equipment Standard Defect Knowledge Base, and the search results have passed verification. This effectively assists maintenance personnel in equipment inspection and maintenance, thereby validating the effectiveness and practicality of the framework established in this research. With the support of domain expert knowledge verification, it can effectively empower related fields to conduct decision support.

First Page

58

Last Page

69

Digital Object Identifier (DOI)

ꎺ 10.19809/j.cnki.kjqbyj.2025.04.006

Download

Share

COinS