Desigh And Implementation Of Complex Emotion Recognition System Based On Knowledge Graph

Emotional intelligence is an important component and development direction of artificial intelligence,which provides a basis for the human-computer interaction system to fully understand human intentions and perform the functions of companionship and comfort.Current sentiment analysis systems usually have problems such as coarse emotional granularity and poor interpretability,which are difficult to meet the application needs of human-computer interaction systems.Therefore,this paper designs and implements a complex emotion perception system based on knowledge graph,and realizes fine-grained and interpretable complex emotion fusion perception based on the high integration of structured data of knowledge graphs and the easy generalization of relational reasoning.The main research work of this paper is as follows:(1)Designed a complex emotion perception system based on knowledge graph,which mainly includes Chinese emotion knowledge graphs,emotional vocabulary analysis and extraction,and complex emotion perception modules.(2)Designed and built a Chinese emotional knowledge graph,which is the basis for textual emotional analysis.The knowledge graph is Based on the Plutchik emotion wheel model and emotion fusion theory.Referring to W3C’s PROV-O and DLUTIR Emotion Ontology,this paper introduces the Seven-step method with ontology evaluation and feedback optimization to develop the ontology,and realizes knowledge representation through OWL language.This paper expands the scale of data using synonyms and realizes the storage and visualization of the knowledge graph using Neo4j graph database.Finally this paper built a knowledge graph covering 10 basic emotions,more than 30 sub-emotions,31441 emotional entities,and 136,161 relationships.(3)Designed and implemented the emotional vocabulary analysis and extraction module,which is the bridge between the human-computer interaction system and the emotional knowledge graph.This paper is based on the Douban Conversation Corpus Chinese multi-round dialogue data set,using BIO mode for sequence labeling,and achieving labeling data expansion through the EDA data enhancement method.Based on the named entity recognition technology based on deep learning,This paper realizes the analysis and extraction of emotional words through the BERT-BiLSTM-CRF model,and compares several commonly used models to verify the effectiveness of this model(4)Designed and implemented a complex emotion perception module,which is the main functional module for emotion analysis of the human-computer interaction system in this paper.Referring to the related theories and algorithms of social network centrality,and based on the PageRank algorithm which measures the importance of web pages,this paper designs a complex emotion fusion algorithm called EmoRank which is based on the knowledge graph.The effectiveness,accuracy and usability of this method are verified by experiments.