Research And Implementation Of Multilevel Sentiment Text Classification System In Chinese

With the development of Internet technology,people getting more and more using the Internet to transfer values and emotions,ordinary messages using high speed Internet spreading in a high-speed.From these messages,we can mining opinions for all things in social,resulting in the work of Internet public sentiment.Internet text can accumulate in a fast speed,in a short time texts over the world can reach to TB.The meaning of sentiment analysis in natural language processing is to determine the text's emotions or opinions with computer technology.It's a very tough task to classifier Internet texts into explicit sentiment orientation,usually the sentiment orientation can be classified as positive and negative,another classification is positive,negative and neutral.Ordinary sentiment text classification in chinese can do nothing to multilevel text,because each of them have a specific research topic.The most popular research topic is weibo short text classification.But ways of weibo short text classification in classify high-granularity text effect is not good.Thus,we present a multilevel sentiment text classification method in chinese,it has the same effect on every level of granularity,and the speed of this method is stable in every kinds of text.The thesis' s main contents include:1.Summary of several sentiment classification algorithm and their basic principle,and discuss the importance and practicability of multilevel sentiment classification2.Compare and distinguish different text granularity with their adopt algorithm.3.Build an useful multilevel sentiment text classification system in chinese,including web crawler module,user management module,sentiment text classification module.For different text granularity,accuracy of system reached 75.8%,feature-1 reached 77.4%,response speed reached 0.2s in local environment and 1.4s in Ali cloud environment.System meet the need of multilevel sentiment text classification in chinese.