Event Extraction And Sentiment Analysis On Mricoblog

With the development of social networking services and thepopularity of smart phones, microblogging has become an importantcommunication and spread flat roof for people, every day there are a largenumber of texts on this platform, which contain a lot of social hotproblems and the emotion information of the users towards these issues. Tofind the topic social points and analyze the user’s sentiment about theseevents is helpful for tackling outburst incident, public opinion survey andmarket analysis, it has a great application value on public opinionmonitoring and marketing management as the high coverage andtimeliness of information on microblogging.Information extraction based on microblogging is a raising researchtopic, including sentiment classification, automatic abstraction, eventextraction etc. The traditional work on event extraction mainly focuses onnews website texts, however, the reproducibility, fragmentation, noisy andrandomness of the microblogging texts makes the previous technology isnot proper for the texts on microblogging.This paper describes an open domain event extraction and sentimentanalysis system based on Chinese microblogging. Firstly, this system willdo a data pretreatment to experimental data, and get the event informationthrough named entity recognition and time resolved technology, thenanalyzes and calculates the proportion about every event to the total events,at last outputs the event which has a large variation of the number againstthe timeline.About sentiment analysis, its main idea is to construct classificationmodel with SVM classifier, the key is the choice about eigenvalue of texts, this paper focus on choosing the network emoticons and words withsubjective emotional as features.The experiments demonstrate that the way this paper proposed forevent extraction and sentiment classification is feasible.