Research And Application Of Sequential Three-way Decisions In Multigranularity Text Sentiment Analysis

With the development of the Internet and social networks,a broad platform has provided for Internet users to interact and express their opinions.A large amount of text data has generated,such as website comments,blog posts,news,etc.However,most of the speeches on the Internet contain expressions of personal emotions,which are not always objective and consistent with the actual situation.Some people freely express their opinions on the Internet and even guide the direction of public opinion,leading a very bad influence.Therefore,the effective sentiment analysis of a large number of subjective texts on the Internet has extremely important research value and practical significance.By analyzing the multigranularity characteristics of text sentiment,this thesis applies sequential three-way decisions theory to the research of text sentiment analysis,which mainly includes the following two aspects:1.In order to correctly classify text,a multigranularity text sentiment analysis method based on sequential three-way decisions is proposed.Firstly,this thesis analyzes the multigranularity characteristics of emotion in text,and propose a multigranularity expression method of sentiment information from coarse to fine according to the amount of sentiment information in text.Then,combined with the idea of sequential three-way decisions,the calculation is gradually executed in different sentiment information granularity and the three-way decisions is carried out for the boundary text.Lastly,the final sentiment classification results are obtained according to the decision cost and threshold setting of different granularitites.The experimental results show that the proposed method has higher classification accuracy and stronger robustness.2.In order to better apply the idea of multigranularity and sequential three-way decisions to text sentiment analysis,a multi-level and multigranularity method is proposed,which is text sentiment analysis method based on weighted multigranularity sequential three-way decisions.Firstly,considering the fact that each granular structure has different weights in actual decisions,this thesis proposes a weighted multigranularity decisions rough sets model with better generalization performance by analyzing the existing multigranularity decisions rough sets model,and its related properties are also discussed.Then,based on the three regions of the weighted multigranularity decisions rough sets model(namely,the positive region,the boundary region and the negative region),a weighted multigranularity three-way decisions rule is obtained.In order to solve complex problems better,a weighted multigranularity sequential three-way decisions model is proposed based on multi-views granular structure and multi-levels granular structure.Lastly,the relationship among granulation cost,granular layer transformation cost and decision cost is discussed.The experimental results show that the model proposed is more efficient and more realistic.