Research On Enterprise Reputation Emotion Classification Model Based On Web Text Mining

Along with the development of the Internet, consumers become more and more interested in sharing about buying tips with all kinds of brand, product and service through blogs, review sites or other channels, and the resulting network word of mouth will directly affect the other consumers’ buying decisions. On the one hand, the good word of mouth can attract more customers, bring greater profits for the enterprise; On the other hand, too much negative word of mouth can reduce enterprise credibility, resulting in the loss of customers. Therefore, Classify the enterprise emotion of network word-of-mouth, analysis of consumers’ emotional tendency, not only help companies respond timely to negative word of mouth, make effective coping strategies; At the same time, through analyzing the fine-grained word-of-mouth text mining, we can found the commercial value, and use them for personalized recommendation of products, apply it to discover user interest, and many other aspects.In this paper, on the basis of Web text mining technology, with sentiment analysis technology as the main line, We studied the key technology of acquisition and preprocessing such as the crawl the Web text data, and the Chinese word segmentation, text stop words filtering; And based on this, we also studied the feature extraction and text vector representation and feature selection impact on sentiment classification; And by improving the HowNet emotional lexicon, based on emotional dictionary’s word of mouth orientation calculation model is constructed, and used in the empirical early word-of-mouth emotion classification; Then build emotion classifiers using K-adjacent algorithm, realization of fine-grained emotion model, finally aimed at same hotel companies for the passion of the fine-grained classification research. In this paper, the main research works include:First, study the DOM tree structure of word-of-mouth HTML, using RostDM software designed for hotel review the URL of the grab and text data acquisition rules, collected a word-of-mouth network in more than six thousand, nearly one million words of hotel reviews as a corpus. The corpus from the consumers’ subjective reviews of hotels, has a professional, obvious advantages, such as emotional characteristics, emotional tendency of research on Web text is have a certain significance.Second, study the effect of feature selection for emotion classification. In the process of training text classifier, feature selection for precision and efficiency of classifier have a significant impact. Based on the K-the most adjacent algorithm training different dimension of feature set, the text classification of emotion, for the characteristics of the training set is not the more the better. Choose proper feature set will help to improve the efficiency of the follow-up study and accuracy.Third, improve the HowNet emotional lexicon, constructed based on the improved emotional bias calculation model dictionary word of mouth. And analyzed the collected more than six thousand public praise in front of the text, and the passion of the slightly at the beginning of classification.Fourth, study the affection of fine-grained classification model, consumers focus on hotel attributes, such as location, service, such as feature selection and weight calculation, and then use emotion tendentiousness of Rapid Miner text mining software, to determine the specific attributes of the emotional tendencies. Fine-grained sentiment analysis to help enterprises more detailed understanding of consumers for their products or services on a single attribute satisfaction, and can better improvement strategies against these attributes.