Emotional Classification Of Movie Criticism Based On Semantic Features

As film and television websites gradually infiltrate people's lives,more and more users will leave their own comments after browsing the contents of the movie,including opinions on the content or a certain aspect and emotional attitudes.Comments on various aspects of the film have become an important information carrier.According to the semantic features of Chinese commentary,such as the combination of part of speech and degree words,emotions are used to evaluate movie reviews using various algorithms and models in the field of natural language processing.Analysis,it is necessary to obtain research on potential emotions.For investors,film commentary has become an important source of data for intelligence analysis.Investors can learn key information such as the viewer's concerns and emotional characteristics,and further determine investment trends or processing methods to meet market and audience needs.For the audience,film reviews significantly influence the choices and decisions of the audience and other audiences.Commenting on movies is a quick way to help other users get relevant opinions in the decision-making process.Researching and analyzing the movie reviews on the web platform can help investors and viewers achieve their goals more quickly and effectively.Therefore,this thesis focuses on film reviews,combining Chinese semantic features and word order relationships,constructing hierarchical conditional random field models and directed network graph models to complete the emotional classification of film reviews.(1)Research on sentiment analysis of film reviews based on conditional random field model.In the machine learning method,the conditional random field model has good classification performance,and.can effectively fuse multiple types of features.The unique semantic rules of Chinese can be transformed into semantic features.The conditional random field model can be used to classify emotions more accurately..In order to solve the problem that the single-layer conditional random field model has low accuracy rate of emotional polarity and intensity classification of film reviews,this paper proposes to construct a three-layer conditional random field model,and complete the subjective sentences and objective sentence classification of film reviews according to layers.Emotional polarity classification of positive and negative comments and emotional intensity classification of general intensity reviews and strong intensity reviews.When emotional classification of film reviews,based on the original semantic features,the effects of transitional conjunctions and progressive conjunctions on emotional polarity classification and intensity classification results are further considered,so that classification features and feature templates are more comprehensive and reliable.Combining the classification advantages of Bayesian classifier and support vector machine classifier,the errors that can be generated by a single classifier are corrected,and the accuracy of classification is further improved.Such hierarchical progressive classification compared with the traditional single-layer conditional random field,through the effective features of each layer of the hierarchical model and the transmission of the feature template to the next layer,can achieve the purpose of reducing classification error and improving overall classification accuracy..Compared with the traditional classification algorithm,the unique semantic features of Chinese can be considered more comprehensively,and the semantic features such as degree words and negative words are integrated into the sentiment classification model to make the performance of the model more excellent.(2)Research on the emotional classification of film reviews with orderly relationship.Aiming at the lack of syntactic analysis and ignoring the rules of word order in traditional research,this paper proposes a constructed directed network graph by combining semantic features such as evaluation words,degree words and negative words selected by conditional random field model and template combination and Chinese word order rules.The model carries out the emotional classification method of movie review.Through the relationship between nodes and nodes in the directed network,the word order relationship of the words in the movie commentary is further studied,and the degree of influence of the semantic features of the ordered order relationship on the sentiment classification is further studied.According to the analysis,the five semantic structures of Q mode,WQ mode,NQ mode,NWQ mode and WNQ mode are obtained.Combined with Weber-Fichner's law,the user's psychological sensation is transformed into physical Quantification,and the evaluation of emotional similarity is better.Determine the emotional polarity and intensity categories of the review.Compared with the traditional classification method for sentiment analysis,combining the semantic structure and constructing the movie commentary directed network graph can fully demonstrate the important role of Chinese lexical order in the emotional classification results,and the different patterns of semantic structure shift the emotional intensity.The degree,the double impact of considering emotional words and semantic structure in the calculation of emotional similarity also has a positive impact on the final classification results.In the case of emotion classification research for film reviews,this paper pays more attention to the unique semantic features and word order rules of Chinese,constructs a three-layer conditional random field model based on semantic features,and builds a directed relationship based on the semantic features selected by this model.Network graph model,such classification algorithms and models can more fully and comprehensively express the emotional characteristics and classification characteristics of movie reviews,effectively achieve the emotional classification of movies,provide assistance for the audience in decision-making,and indicate to investors determine the investment trends and directions.