A Deep Learning-Based Entity-Relationship Extraction Method in the Field of Electric Power Public Opinion

Abstract

Entity-relationship extraction can obtain key information elements from texts. Electricity opinion texts have the characteristics of complex entity relationships and less annotated data, so it is difficult to find entity information with mutual relationships from text data in the field of electricity opinion. To solve the above problems, we propose the ABAC (ALBERT-BiLSTM-ATT-CRF) model to extract entity relationships in electric power opinion texts. By using the pre-training model of ALBERT and combining the five-stroke sequence, radicals and pinyin of Chinese characters to extract features, and the features of these parts are fused to enhance the ability of extracting text feature vectors. The experimental results show that the accuracy of entity-relationship extraction has been significantly improved, which verifies the effectiveness of the model designed in this paper for entity-relationship extraction in the field of electric power public opinion.