Research On Key Techniques For Subjectivity Detection Of Microblogs

With the development of network, Microblog has become a cross time product.Microblog is a user-relationship based platform to assist user sharing and gaininginformation. Via various client tools such as WEB and WAP, users are able to create shortmessages in less than140characters. As Microblog booms, microtext is made large scale.The research on the microtext has thus become an important topic.This thesis concentrates on key technology to detect subjectivity of Microblog. Thecontributions of this works are summarized as follows:A Thread-based Two-stage Clustering Method of Microblog Content TopicDetectionBased on the features of Microblog texts, such as short, semi-structured, contextdependent, we propose a thread-based two-stage clustering method. In first phase, thetemporal-author-topic (TAT) model is applied to clean the thread, namely to filter noisyMicroblog texts out of each thread. In second phrase, Microblog texts with each thread aremerged to form the thread text so that the TAT model is applied to find global topics. Thisapproach release the data sparseness problem effectively in Microblog.Using Cross-Entity Inference to Improve Event ExtractionWe regard the entity consistence as an important feature in event extraction, therelationship between entity type and event type is considered, the entity type can be used topredict event type, then we choose features to construct classifiers, which are utilized torecognize elements of an event. Compared to traditional sentence-level approaches, thismethod can acquire better effect.A Grammar-based Unsupervised Method of Mining Volitive WordsWe adopt the method based on2-gram,3-gram,4-gram, to extract the Chinesevolitive words. If the event extracted from the Microblog contains volitive words, it will be identified that this Microblog has subjectivity. The proposed approach mainly includesPOS rules and n-gram features, and it consists of two phases, the second phase is iterationof the first one.