Event - Based Quadripartite Representation of The Power Operation Text

Abstract

The growing maturity of the Power Grid industry and the continuous progress of information technology make big data mining in the Power Grid industry possible. There is no standardized requirement for the description of operation and maintenance data in the Power Grid industry, resulting in a large amount of unstructured text data. Aiming at solving this problem, this paper analyzes the characteristics of the unstructured text data and addresses the importance of text framework. Then, the event representation framework and event slot are defined, and the event elements are extracted using the methods of part-of-speech segmentation, semantic dependency analysis and dependency syntactic analysis in natural language processing(NLP). Finally, the event quaternion construction method is given. In this paper, the feature points of operation and maintenance data are found and summarized as ‘two events and four parts'. The event slots and event quaternions are defined to successfully structure the unstructured text. It provides the possibility for operation and maintenance data to be applied to question and answer system, intelligent order distribution, spare parts estimation.