Sentiment Analysis Of Short Comments On Animated Films Based On Text Mining

In recent years,more and more excellent domestic animated films have risen,but there is still a gap compared with the United States,Japan and other developed countries.But how to change the current situation and improve the status of domestic animated films,in addition to qualitative analysis,there are documents that excavate the deep views of the audience through quantitative analysis of the text on the basis of animated film review,which has important practical guiding significance for the development of animated films in the future.This paper attempts to further explore the modeling and analysis methods of animation film reviews in this direction,and tap the audience’s views on animation films in different countries and the shortcomings of domestic animation films.After consulting a large number of literatures,it is found that the number of lowfrequency words in the corpus accounts for a high proportion,but its importance is generally low.Therefore,combined with the Pareto principle,this paper proposes a method of dynamically constructing stop words based on the word frequency distribution in the corpus.The core of this method is to determine the low word frequency based on Zipf s law,according to the short text characteristics of animation film review,and incorporate it into the stop list which can be dynamically adjusted according to the corpus.Then combined with LDA theme model,Zipf-LDA model is established.Since box office is one of the most important economic indicators for evaluating films,in order to comply with the future trend,this paper mainly focuses on the douban short comment data of the top 50 high box office animation films in the mainland.Based on Python software,this paper applies Zipf-LDA model to empirically analyze the sample data,and compares the general model with confusion as an indicator;Combined with K-means clustering,the film elements behind film reviews are further mined.The empirical results show that:(1)Compared with the general model,Zipf-LDA theme model is better,it greatly reduces the degree of confusion,which can optimize the theme modeling results of online comment texts and improve the running speed of the model.(2)Further,this paper obtains the theme keywords through K-means clustering,which can be summarized into three categories:story plot,film theme and film details.(3)According to the comparative analysis between countries,compared with the United States and Japan,domestic animation films have deficiencies in original IP,production companies and film details.According to these conclusions,this paper puts forward relevant suggestions from the micro and macro levels,which provides a reference for the development of China’s animation industry.