Research on the discrepancy distinguishment of the electric power standard clauses based on NLP
Qian Zheng, Yibo Yong, Jiajun Lin, Fei Xu, Fu Hui, Zhenxu Sun · 2022
To solve the problem of overlapping, repeated and contradictory terms in different kinds of electric power standards, we propose a method of discrepancy distinguishment for the electric power standard clauses based on natural language processing (NLP). The proposed method converts the unstructured text into triples using the entity relation extraction method for power standard terms, and stores them in knowledge graph to reduce the query time and the processing time. Then, the text similarity calculation method is used to compute the similarity of triples. Finally, the discrepancy distinguishment of standard terms can be obtained. Taking power grid equipment technical standard difference clause data as an example, experiments were carried out, and the recall and the precision reached 82.2% and 70.6% respectively. The experimental results illustrate that the recognition rate and adoption rate reached 82.2% and 70.6% respectively. It also shows that the proposed method could provide a solution to the discrepancy distinguishment of standard documents in various industries.