Research On Emotional Classification Of Online Movie Reviews Based On SOW-BTM

With the rapid development of Internet information,more and more users tend to express their views and attitudes on the Internet public platform.Based on this,sentiment classification has also been applied to mining subjective information of movie reviews.By emotionally classifying movie reviews,not only can the user be provided with the overall emotional tendencies implicit in movie reviews and assist users in making decisions and satisfy their personalized viewing requirements,but also can provide feedback for film distributors.At the same time,It can also be effectively monitored for social media and give the policy support for government.It can be seen that it is of great significance to analyze the potential value of the movie reviews by analyzing the user's emotional tendencies.However,the current film reviews are not only large,but also short and trivial.And the traditional emotion classification method can't handle the specific field of emotional information well.It poses a great challenge to the classification of emotions in the field of film reviews.Based on the problem,Aiming at the special field of film review,this paper proposes an unsupervised sentiment classification method based on the SOW-BTM model for short textual film reviews.The contents of this paper are as follows:(1)In response to the limitations of traditional sentiment classification methods in the field of emotional problems and the use of short texts,the BTM topic mining model was improved and SO-PMI was used as a weight model to measure the mutual information between vocabulary and emotional seed words,different vocabularies are given different weights in Gibbs sampling.Changing the probability distributions of "document-subject" and "subject-vocabulary" in the traditional BTM model to the probability distributions of "document-subject" and "topic-emotional word" by increasing the weight of emotional words and reducing the weight of non-emotional words.Finally,forming a SOW-BTM model based on the three-tiered Bayesian structure of "document-subject-emotion word".the BTM topic mining model is more appropriately applied to the field of sentiment classification by the improvement.(2)On the basis of the data obtained in the field of film,Creating an emotional dictionary that is applied to movie reviews by combining the semantic similarity calculation based on Hownet with a self built dictionary.The emotional dictionary get rid of the limitations of the traditional emotional dictionary can not identify the domainemotion words.By using the emotional dictionary to calculate the emotion value of emotional word under the theme keywords,to get the emotional value of the topic,finally calculating the emotional tendencies of the document,then improving the accuracy of field sentiment classification.(3)Apply the SOW-BTM model to the emotional classification of movie reviews.The data set used in the experiment came from the comments of three films crawling in the Douban.com.The number of words in each comment is not more than 150.Finally,through these comments,verified and compared the application of SOW-BTM model,BTM model and LDA model in emotional classification from two aspects of horizontal and vertical.The results show that the SOW-BTM model is superior to the common BTM model and the LDA model in the emotional classification of short text movie reviews.