Research On Sentiment Classification Based On Semantic Lexicon Of Hotel Field

In this paper the related theory and technology method of Chinese emotion classifiers areorganized, and based on the hotel field to build the emotional lexicon. On this basis, supportvector machine (SVM) method is applied to build the emotional classifiers, and build aclassifier comparative experiment with the emotional intensity value of the emotionaldictionary. The main work is as follows:(1) Using of statistical and machine learning to mining the database, and based on therelevant dictionary of Hownet to build a special emotional lexicon in the field of hotel. Innetwork terms and hotel field corpus expand the Hownet dictionary, to form a relativelycomplete emotional dictionary of hotel field. At the same time, this article also referenceexisting modify dictionary, as a correction of emotional dictionary. And give thecorresponding emotional intensity values to the words in the two dictionary.(2) Based on emotional dictionary and support vector machine (SVM) theory buildingtext classifier, and constructing classifier based on emotional dictionary emotional intensityvalue. This article uses the the support vector machine (SVM) text emotion classifiers toclassify the critical text in the field of hotel, critical text can be divided into positive andnegative emotional categories. At the same time, using the emotional intensity valuesclassifier to classify the critical text, critical text can be divided into positive, negative andneutral emotional categories.(3) The experiment and comparative study of the two kinds of classifier. In the process oftext classification, different feature selection methods will produce certain impact to textclassification effect, this paper select document frequency,χ2inspection, the emotionaldictionary of hotel field and dictionary of Hownet, experimental results on four differentkinds of feature selection for analysis and comparison, the experimental results prove that theemotional dictionary in the hotel field of support vector machine classifier in recall, precisionand micro average has certain advantages.The two different emotional classifiers experiment in the field of hotel comment text,found that the accuracy difference is not big, preliminary verify the feasibility of the twomethods.