Research On Psychological Early-warning Model Based Fine-grained Sentiment Lexicon

Along with the rapid development of the economy globalization,college students not only face the opportunity of increasing social demands,but also bear the competitive pressure.In recent years,more and more events with very negative effects have been brought about by psychology fault.However,psychology teachers are in short demand today,they can not check every student and take precautionary measures.Therefore,the psychological crisis early-warning model research is a crucial work.Nowadays,the development of the Internet bring a new broad of perspective on this study.As the new media of spreading information,micro blog has become the important method for the flint in to obtain knowledge and information.This thesis aims to extract information from vast irregular dataset,and explore the user's emotional state change trend.Nevertheless,micro blog content allows user send short documents,and the expression of users is fairly vary differently.Each problem presents challenges to the study.The thesis discuss these questions mainly from the following aspects:1.Construct a fully functional fine-grained sentiment lexicon: Sentiment lexicon is the core of sentiment analyse.This thesis make full use of both Chinese and English bilingual knowledge to build a more complete emotional dictionary.2.Algorithm of emotion tendency computing based on weight factor: Negativity and degree adverbs are two kinds of emotion factor influencing emotional vocabulary orientation,these two kinds of emotional factors show a different impact on emotional words in different position.3.Build a psychological early-warning model based on fine-grained sentiment lexicon: This thesis builds a psychological early-warning model combined with the emotional dictionary and text emotional algorithm.Experiments show that the psychological early-warning model has achieved good results.