Predicting Relations in SG-CIM Model Based on Graph Structure and Semantic Information
Pengyu Zhang, Yufei Li, Xinliang Ge, Wenhui Hu, Lizhuang Sun, Xueyang Liu · 2022
State Grid Enterprise Common Data Model (SG-CIM) is a semantically unified data model for smart grid business applications. The model is divided into entity and relation files. Since the relations in SG-CIM model are manually labeled, which consumes a lot of labor costs and is also prone to omission and mislabeling. In order to solve these problems, this paper introduces a method for predicting complex relations in SG-CIM model based on graph structure and semantic information. The relations of graph structure and semantic information are extracted and used by BERT and ERNIE models to predict the relation types between head and tail entities. In order to improve the efficiency of relation prediction, the five relation types are classified into the parallel type and the inheritance type based on the structural relation between entities, and the pairs of entities are separately mined using frequent term mining and K-nearest neighbor methods. Using these methods for predicting complex relations in SG-CIM models can effectively save labor costs.