Analysis Micro-blog Emotional Tendency Based On Nuclear Sentences And Syntactic Parsing

With the flourish of Chinese micro-blog in recent years, people prefer micro-blog to give their opinions, show personalities and express emotions. Hence sentiment analysis of micro-blog has become a hotspot in Chinese information processing field recently. Researchers apply the evaluation objects and sentiment orientation, gotten from the text of micro-blog by the means of machine learning and statistical methods etc, to customerrelationship management, micro-blog marketing, product reviews analysis, branding promotion and public opinion monitoring.Through sentiment analysis research on Sina micro-blog, this paper puts forward a new micro-blog sentiment orientation analysis based on context dependency relation, that’s fractional steps. By method of fractional steps, firstly, find out text evaluation object, then find out effective concerning sentiment words through syntactic analysis from evaluation objects, in this way, it reduces interference from irrelevant sentiment orientation words, which makes it more correct when judging sentiment orientation.Besides, this paper analyzes the characteristics of micro-blog, including expressions, word-using habits, punctuation and so on. According to these characteristics, we can summarize rules how to extract main components of sentiment expressions, and then extract main part of sentiment orientation expression. As experiment turns out, the accuracy of extraction based on sentiment evaluation objects and effective sentiment evaluation words is much better than extraction directly from original sentence.Conditional Random Fields Model is adopted to extract evaluation objects in this paper. Besides, this paper bring into dependency relations, relative position, semantic information as extended features except word in the model training, which has a profound effect.Syntactic analyzer is used to analyze kernel sentence dependency relations to extract effective sentiment expressions of evaluation objects, and then set various weights value according to different relative positions.In the end, calculate the sentiment orientation of sentence by the use of sentiment dictionary. As experiment turns out, it is better to judge sentiment orientation based on accurately got evaluation objects rather than direct sentiment orientation analysis on micro-blog.