Research On Fine-grained Sentiment Analysis Of Uyghur Text

Text sentiment analysis has attracted more and more attention in recent years, and it is research hotspot in the field of information extraction, data mining, natural language processing and so on. With the rapid development of internet technology and especially with the rising popularity of the Web 2.0, the large number of Uyghur websites, social platforms and the platforms of business application emerge as the times require. There is great significance of research and application in studying fine-grained sentiment analysis for Uyghur text. The traditional research methods of sentiment analysis are largely to determine the emotional polarity of word, sentence or writing. Although the traditional research methods have obtained the certain achievement, they ignore the topic that sentiment expressed and there is less study on fine-grained sentiment analysis at claim level. In addition, most of these research methods are only able to analyze the opinion which can be expressed clearly in the subjective text, but can’t analyze the opinion that can be expressed implicitly in the text.This paper sets Uyghur texts as the object and studies the method of fine-grained sentiment analysis, and it includes the following two aspects:(1) Research on sentiment analysis at claim level. We propose a method based on cascade CRFs model to analyze the sentiment at claim level of Uyghur opinioned text, which combines the unique sentence characteristics of Uyghur text. The first layer extract s the topic word and its corresponding opinion word, and determines the scope of opinioned claim, then the result is passed to the second layer as one of the key features which contributes to sentiment analysis at the claim level. The goal of the sentiment analysis on fine-grained opinion mining is to build a quadruple, which is <opinioned claim, topic word, opinion word, sentiment>.(2) Research on implicit sentiment analysis. The morphology, idioms, the punctuation, the word depending on context, MI and negation are employed to analyze the implicit sentiment. The MI can compute the dependence of the context information. We construct the new model to study the implicit sentiment for Uyghur text, and the model is based on CRFs. The imbalanced dataset of the corpus will cause the non-ideal result in analyzing implicit sentiment. In order to improve the situation we improve information gain(IG) which is the traditional feature selection approach when conducting feature selection.The experimental results show that the cascade CRFs model applies to sentiment analysis at claim level for Uyghur text and its effectiveness has been reflected. The analysis of implicit sentiment promotes the further study on the task which is fine-grained sentiment analysis for Uyghur text.