Disambiguating Sentiment Ambiguous Adjectives And Its Application Study

With the Internet, WEB2.0and e-commerce development, microblogging, stickers and website comments, a lot of people for goods, services, news, ideas, experience and opinions. Accurate analysis of user comments to express the emotional tendency is to appraise the task of analysis. Sentiment analysis is divided into tendentious analysis of the chapter, sentence and word level. Word level is the smallest classification unit is the most basic classification unit, among these terms, sentiment ambiguous adjectives occupies a considerable portion. Sentiment ambiguous adjectives is an inseparable component to appraise the analysis, but on sentiment ambiguous adjectives the studies was few, extracted from the sentence appraise the collocation method to constitute a collocation table, and improve the rule base and appraise vocabulary library, improve the accuracy rate.Classification of sentiment ambiguous adjectives is the main key technologies of the two categories are based on statistical methods and methods based on the dictionary rules and based on the statistical method of maximum entropy, Bayesian classifier, KNN, Support Vector Machine, mutual information, and other mathematical and statistical models, feature extraction, classification of sentiment ambiguous adjectives appraise tendentious; and vocabulary-based approach mainly in the extract feature words, and then compare the classification based on HowNet, WordNet, Chinese sentiment word table, synonymous with the word Lin et al, to determine the tendency of the sentiment ambiguous adjectives appraise.In this paper, based on collocation table to classify the sentiment ambiguous adjectives and applied to s the process of analyzing the corpus syntactic analysis, extract collocation with the word, constructed with the Collocation table, and then use the collocation table to determine sentiment tendency of the multi-polar adjective to distinguish the sentiment ambiguous adjectives appraise tendentious, and finally will extract the sentiment ambiguous adjectives appraise marked tendency to added to the text analysis, the final text appraise classification results. In this way, effectively make the multi-polarity of adjectives in text categorization blind, sentiment ambiguous adjectives to the text orientation analysis, improve the quality of the classification.Collocation to determine appraise sentiment ambiguous adjectives has a strong context-sensitive, so need to extract collocation, this article is based on the hotel evaluation classes, IT classes and competitions evaluation of multi-domain corpus, the use of syntactic analysis to extract with words, by calculating appraise frequency will be paired with the word classified, and ultimately the accuracy of classification of multi-polarity of the adjective.In the experiment, multi-domain corpus and the competition to a single field of hotel evaluation corpus as test corpus. Experimental results show that multi-polarity of the adjective marked by this method to appraise the accuracy of the tendentious in a single field of corpus reached95.43%,95.13%in the interdisciplinary corpus. Meanwhile, Sentiment ambiguous adjectives added to the emotional, emotion analysis system to appraise the accuracy of classification by93.63%to93.81%.