Research on the Method of Policy Text Knowledge Mining Based on Neural Network

Abstract

AbstractIn view of the large amount of information in domestic policy texts and the rich semantic information, it is difficult for researchers to quickly and effectively sort out and compare and analyze the content when reading. This paper takes the policy text data resources in the energy field as an example, and sorts them out according to the types of resources. On the basis of sorting out the policy resources, it combines the characteristics of policy text information with deep learning technology, explores new ways and methods to interpret policy text content in a deep learning environment, and further evaluates and analyzes the effect of the neural network models for entity recognition and relationship extraction of domestic policy texts. This paper uses Bi-LSTM to extract text features in both directions to solve problems such as long-distance dependence, introduces the attention mechanism to obtain important features, and effectively reveals various entities and concepts in the policy text and their relationships, thereby assisting researchers in better policy implementation in-depth interpretation and analysis of text content.KeywordsDeep learningPolicy textNeural networkEntity recognitionRelationship classification