Key Technology Research And Implementation On Emotion Analysis Based On Micro-blog Text

With the development of social network and its self-media such as micro-blog, thereusually appears almost hundreds of millions of micro-blog text everyday. As the massmicro-blog text usually contains some kinds of information about personal, social,enterprise in multi-dimensional, multi-level and diversification style, it is necessary toanalyze the content, to monitor the network public opinion and to analyze thecorresponding sentimental tendency. The research has important values both in theory andapplication.This thesis presents an algorithm on collecting micro-blog data based on simulatingbrowsers’ behaviors. By using some natural language processing approaches such assegmentation, part of speech tagging, keyword extraction, and on the basis of thecorresponding sentimental base and micro-blog chunks, we present approaches on settingvector space model and adjusting weights of sentimental influence factors dynamically,and then do micro-blog content based sentimental analysis. The main research works areas follows: First, on the basis of simulating browsers’ behaviors and the correspondingtools such as HttpWatch8.5, we collect the mass micro-blog data. Second, on the basis ofHidden Markov Model and N-Gram language model, we present an effectively Chinesesegmentation tools named as SkyLightAnalyzer, which can do the segmentation, part-ofspeech tagging, word sense disambiguation, and unknown words recognition. Third, onthe basis of the combination of statistics and rule based algorithms, we do the keywordand sentimemtal unit extraction on the basis of the above segmentation. Fourthly, on thebasis of setting vector space model and adjusting weights of sentimental influence factorsdynamically, this thesis presents the blogger personal modelling and content analysisbased micro-blog sentimental analysis. The experimental results and the analysis show thefeasible of the approach. Further works are also present in the end.