Research On Key Technology Of Sentiment Analysis For Chinese Language

With the development of information technology,more and more information appears in the internet with the form of unstructured text,such as a large number of subjective texts in forum,blog,micro-blog and so on.Therefore,sentiment analysis has become a hot research topic in the field of natural language processing,and it has been widely applied in public opinion monitoring,commodity recommendation,comments analysis.In this paper,we take the field text(financial and stock market)as research object and our goal is to improve emotional words recognition accuracy and recall rate.The main work is as follows:1.Dictionary research aspect.First we crawl financial stocks corpus from relevant forum and artificial-mark emotional words.Second we combine these emotional words and existing emotion dictionary to get basic emotion dictionary.Final we expand the basic emotion dictionary then get financial emotion dictionary.The dictionary includes nine category words such as:positive emotional words,negative emotional words,transitional words,coordinating conjunction words,negative words,degree words,named entity,unit quantifier,new words.etc.2.Algorithm research aspect.We innovatively sum up the emotional pattern sets and POS pattern set,then based on these two pattern set we propose POS check algorithm and emotional pattern matching algorithm.The former is used to verify or tag POS.The latter is used to automatically discovers new emotional words,then improves the accuracy of emotion recognition with adding the new emotional words to sentiment dictionary.3.System realize aspect.We development chinese sentiment analysis software system and integrate related function module,including architecture design,algorithm realization,data visualization,software GUI design.Finally,we analysis NLP toolkits in current stage.This paper takes NLPIR as segmentation tool,and in order to improve the accuracy of specialized field segmentation and POS tagging we import emotion dictionary as user dictionary.4.Experiment aspect.The first experiment is tagging POS of wrong-segmentation word,we select 300 sentences which contain 111 test words.Finally,compared to artificial-mark words,experimental result shows the algorithm has a good accuracy.The second experiment takes sentences with transition and coordination rule as object and uses traditional dictionary matching algorithm and emotional pattern matching algorithm to recognize sentiment word.Finally,taking artificial marked result as standard,the experiment shows emotional pattern matching algorithm has a better matching result compared to dictionary matching algorithm.