Research On Sentiment Analysis Method Of Tibetan Text Based On Big Data

This paper selects Tibetan news texts from the massive data information as the research object.The process of sentiment analysis of Tibetan news texts is divided into the process of crawling and preprocessing of corpus,the construction of basic emotional word dictionary,the expansion of emotional word dictionary,and emotional calculation.The theoretical basis of each stage is proposed.Research methods and design corresponding experiments to achieve and verify.main tasks as follows:1.In the collection of Tibetan corpus,firstly use the reptile technology to collect large-scale Tibetan text information from Chinese and foreign news websites such as China Netcom,People's Daily,Xinhuanet,etc.,and then reduce the noise of the collected information,and finally establish A large-scale Tibetan news corpus for Tibetan sentiment analysis.In the construction of the basic sentiment word dictionary,aiming at the lack of the emotional word dictionary,based on the existing work of the laboratory,the construction of the emotional word dictionary is carried out,and the method of using the word vector to expand the emotional word dictionary is used to collect the news website.The large-scale Tibetan text is processed,and then the emotional words are automatically extracted from it,and a more practical Tibetan emotional word dictionary is established.In the Tibetan chapter-level sentiment analysis,firstly,based on the method of emotional word dictionary,the large-scale Tibetan news corpus for Tibetan sentiment analysis is automatically labeled,and the SVM model is constructed by using the corpus,using dictionary,SVM,dictionary + SVM,dictionary + word vector + SVM and other methods training model,emotional analysis and sentiment orientation analysis of large-scale Tibetan texts collected by orientation,through experimental comparison,dictionary + word vector + SVM method training model A good result has been obtained.