Intelligent Analysis of Differences in Electricity Industry Standards Through Natural Language Processing
Year: 2024
Type: book-chapter
Abstract: This paper presents an intelligent method for analyzing differences in the electricity industry standards by utilizing natural language processing. The main goal of this approach is to detect variations between standards and identify their distinguishing factors, aiming to enhance standardization, regulation, and consistency. Consequently, it will improve the safety and reliability of electricity equipment and systems while reducing production, operation, and management costs. To begin with, the paper constructs a dataset of electricity standard differences based on the open-world assumption theory. Then, it employs a noisy technique to fine-tune the SBERT model, enabling the identification of discrepancies in electricity standard clauses. Lastly, by optimizing the SimCSE model using a relaxed optimal transport distance, the interpretability of the model is enhanced, and a matrix representing text similarity is generated. This matrix facilitates the visualization of discrepancies in clause text. The proposed method achieves precision and recall rates of 81.26% and 82.90%, respectively, for identifying standard discrepancies. By employing this method, not only can the sustainability of the electricity industry be improved, but it also provides valuable data and decision-making references for electricity enterprises to effectively address issues related to standard management and implementation.
DOI: 10.3233/faia231412
Authors: Xiaoxin Gao, Yongqing Liu, Ming Sun, Wangning Guan
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4391753676