Sentiment Analysis In Chinese Thinking Model

Sentiment Analysis is an important part of the Emotion Computing. Based on researches on sentiment analysis of words, sentences and the differences between Chinese thinking model and Western thinking model,we proposed a method, which was a sort of lexicon-based method, which could attain the emotion of articles.To accomplish COAE2009's(Chinese Opinion Analysis Evaluation) task one and task two, we did some researches on the words and sentences which are the basic elements of an article from scratch. Meanwhile we took full advantage of Dalian University of Technology Affective Lexicon Ontology to distinguish the emotional words from others and discerned the emotion of a sentence by the emotional words, adverbs, negative words, sentence length and but-clause.The evaluational result of our achievement shows that our method gave the priority to strong-emotion sentences, and achieved an advance that is higher than average P@1000 by 16 percents.During the researching of sentiment analysis on articles, we discovered that the factors which could influence the emotions of articles were not only the decorative elements but also the thinking model of people, which were not paid attention by other researchers. We proposed a set of methods to compute the Chinese thinking features such as graphical model, specific model and scattering model compared with Western thinking models such as straight model, abstract model and focusing model.As to the weak-emotion sentences,we use LSA methods to clarify the emotion of uncertain samples by retrieving the seed samples in the semantic space. The experimental results showed that the performance of sentiment analysis involved with thinking models was much better than the lexicon-based methods.So we could make a conclusion that thinking model was an important factor in researches on sentiment analysis and our method was an effective method to the sentiment analysis.