A Study Of The Expression Of Deep Relations In Emotions And Affective Computing

With the development of information technology,a large amount of textual information of a review nature has been accumulated on the Internet.These texts contain the rich emotional expressions of reviewers,which has contributed to the development of natural language processing.In natural language processing,semantic relations play an important role.Obtaining effective semantic vocabulary and semantic relationships from massive information is conducive to the accuracy of text sentiment analysis.This paper mainly studies the lexical extraction of Sentiment semantic relations,the method of Sentiment semantic similarity and the calculation of emotional weights,and realizes the multi-classification of text sentiment.The main tasks include:(1)The method of semantic relation extraction is studied.The words with semantic semantics were extracted including synonymous relations,antonymous relations and behavioral semantic relations.Based on the "Modern Chinese Dictionary" and the vast amount of information on the Internet,define rules to extract synonymous and antonymous relationships,and then use the part-of-speech tagging to extract behavioral semantic relationships.This paper use a method of random selection and manual verification for evaluation.The accuracy of synonym extraction is 96.25%,and the accuracy of antisense relationship extraction is 97.43%.(2)The calculation method of semantic similarity is studied.The concept of semantic similarity is introduced,and the rules are defined based on the Modern Chinese Dictionary and the vast amount of information on the Internet.According to the different rules and recursive times,the calculation method of semantic similarity of affective words is defined,and this method is applied.(3)The calculation method of emotion weight is studied.Construct a degree adverb dictionary and divide the degree adverb level to calculate the emotional weight of the vocabulary.Emotion words with different emotion weights are extracted according to this method.However,due to the limitation of rules,not many words are extracted.(4)Emotional computing based on deep learning is implemented.Analyze and summarize the existing methods and effects of sentiment analysis.Multi-classification of text sentiment is realized.