Text Sentiment Classification Based On Dynamic Benchmark

The frequent developing of Internet extensively gives people aplatform on which people release emotion and give personal opinionslaxatively. Faced these emerging online comment, government,manufacturer and consumer wish to know what attitude people havetoward public opinion, commodity or service. It’s difficult or almostimpossible to organize and collect their text sentiment orientation relyingon manual work. So, it is urgent to have a text analysis technology whichcan make out emotion and attitude of comment texts automatically，namely technology of text orientation analysis, that is called OpinionMining.Aiming at such online comment texts, this paper analyzes theorientation at three levels: word, sentence and text. Firstly, Aiming at theproblem that semantics orientation calculation of word based on fixedbenchmark excessively dependent on benchmark words, based on thepart-speech, this paper suggests the algorithm of semantics orientationcalculation of word based on dynamic benchmark. Then, build the dualdynamic benchmark----field benchmark, calculate semantic orientationcalculation of polar words in field text, explain the method of text orientation analysis based on field benchmark. Lastly, the paper proposesEmotion Submit ES algorithm to fulfill the topic mining and topicsentiment classification in sentence and Subject Emotion Count SECstrategy to implement the analysis and judgment of the whole commenttext based on sentence structural analysis.Experimental results indicate the advantage of text sentimentclassification based on sentence structural analysis which proceed fromsentence topic; The recall rate in experimental results show that ESalgorithm are sensitive to the comment text with emotion orientation.Meanwhile, high precision validates the validity and feasibility of thealgorithm. Furthermore, high stability which the method is shownillustrate that text sentiment classification based on sentence structuralanalysis can be widely used in other fields.