Research On Quantitative Analysis Of Fuzzy Semantic In Chinese Emotional Words And Its Application

With the rapid development of Internet technology, Internet has become an important source and a platform for people to obtain information or express personal opinions. How to use the various network comment information to analysis the reviewers’ emotional tendencies effectively, has become one of the hotspots in the field of Natural Language Processing. Text sentiment analysis determines the emotional polarity and extent of text by mining and analyzing the position view, emotional and other subjective information, which has broad application prospects in product tracking and quality evaluation and has important practical value in the field of information retrieval, online public opinion risk warning. As the emotional word is the essential research object in text sentiment analysis, this paper aims to research the quantitation and application of fuzzy semantic in Chinese emotional words. The main work lists as below:1. Divide Chinese sentiment words into fundamental sentiment words and compound sentiment words based on analyzing the characteristics of Chinese language and Chinese emotional words. According to the structural features of fundamental sentiment words themselves, this paper divided them to study the fuzzy semantics.2. On fundamental sentiment words, propose a method of polarity computation of Chinese sentiment words based on Gaussian distribution. This method tends to get the emotional value of each word from a range, which can correct error from statistical calculation. Experimental results show that, the quantization accuracy of fuzzy semantic in fundamental sentiment words improved in the respective error range. When the error interval was shortened to 0.1, the accuracy was increased by 7.6%.3. On compound sentiment words, propose a multi-strategies sentiment tendency quantitative method. For the seven structural of compound emotional words, the appropriate quantization strategies are designed. Experimental results show that the quantization accuracy of fuzzy semantic in compound sentiment words improved 4.4% when the error range is shortened to 0.1.4. Proposed a fusion text sentiment analysis method which based on SVM, NB and the above methods. This method takes into account the adversative conjunctions impact to the sentence emotional. The experimental results which test in the real hotel reviews corpus show the accuracy of polarity judgment increased about 4%. For all, the former methods were proved effective again.