TS-BP Mixed Sentiment Computing Model Based On Domain-specific Sentiment Dictionary In The Field Of Chinese Movie Review

With the rapid development of Internet technology,more and more people tend to express their views and feelings on the network platform.Online movie review has been developed gradually in this background,which has attracted the attention of many scholars.Movie review as the audience's evaluation of the movie,can directly reflect their feelings of movies,its sentiment analysis has important application value,such as movie screening-oriented for cinema,box office prediction.Compared with the English movie review sentiment analysis,domain-specific sentiment dictionary and tool that can be used in the field of Chinese movie review is relative less.But Chinese movie review has more complex semantics,so,there are fewer scholars studying in the field of Chinese movie review sentiment analysis.In this paper,TS-BP(Topic Sentiment Back Propagation Model)mixed sentiment computing model is proposed based on movie review domain-specific sentiment dictionary(MRSD),considering the thematic factor and the latent semantic relations among the topics.The main works are as follows:(1)The selection of seed word is the basis of the construction of MRSD.In this paper,the specific implementation of seed word selection based on semantic similarity and improved word clustering algorithm is given.The improved word clustering algorithm not only takes into account the semantic relationship between words and words,but also takes into account the contextual relationship between words and words in the document.In this study,the best practice value of the number of seed words is given through experiment.The result shows that it is better to construct the domain-specific sentiment dictionary by using the seed word selection method proposed in this paper.(2)In the face of the dilemma of lacking resources in the field of movie review domain-specific sentiment dictionary,this paper builds MRSD by using the basic sentiment dictionary and the improved SO-PMI algorithm,and improves the movie review word's recognition accuracy of sentiment polarity.The results of the contrastive experiments show that the sentiment dictionary constructed in this paper has good effect on the movie review sentiment analysis.(3)The emotions expressed by the movie review are mainly directed at some of the topics in the film,and the influence of the sentiment words on the sentiment computing is different under different topics.In this paper,the topic and sentiment words are encapsulated into an entity,and the concept of topic-sentiment(T-S)entity is proposed.Based on different topic extraction methods,the T-S entity construction algorithm is given.Considering T-S entity context,topic weight and other factors,the T-S entity sentiment calculation method is given based on the MRSD in this paper.The experimental result proves that the T-S entity computing model is reasonable and effective in the analysis of Chinese movie review.(4)Considering the potential semantic relations among the different topics,this paper presents the vectorization scheme of movie review based on T-S entity,and proposes a TS-BP mixed sentiment computing model.TS-BP mixed sentiment computing model considers the influence of the topic on the sentiment computing and the potential semantic relations among the topics,and improves the overall performance of Chinese movie review sentiment analysis.The experimental results show that the TS-BP mixed sentiment computing model has an average accuracy of 94.7%,which is higher than that of other movie sentiment computing methods,and it shows the validity of TS-BP mixed sentiment computing model proposed in this paper.