Research On Online Multimedia Microblog Topic Mining Algorithm Based On Comments

With the development of the web 2.0, the form of data from Internet are more enriched, and these data contain a great value. Recently, A large number of researchers are engaged in text mining research from the huge web data. Microblog is a kind of information interaction platform with rich multimedia information, and has a profound influence among its' large user community. Usually the text content of a microblog is short and the topics are usually embedded in images or videos and other multimedia content, so how to mine the topic of complex multimedia information and use a concise and effective text model to express the topic of the multimedia has important research significance.However, the existing text mining model and method cannot effectively mine and fully display the topic of multimedia microblog, this paper proposes a new text description model and topic mining algorithm.This paper first studies topic detection and tracking theory, topic tracking technology and the topic model theory, and collect the related theory from existing theory as a foundation in this paper.For supervising the topic mining based on comments, this paper introduces the feature of microblog topic heat dynamic evolution and topic content dynamic evolution, and put forward Microblog Online LDA(MBO-LDA for short) model based on the online LDA models, MBO-LDA is used to model online microblog text stream.Then based on MBO-LDA model, this paper designs the online multimedia weibo topic mining algorithm based on comments, and use the Multimedia Microblog Text Description Model with two dimensions of content and emotion to show the result of topic mining. At last, this study conducts a series of contrast experiment with experimental data built by216345 Sina Weibo that collected by Sina Weibo crawler tools according to publish time. Experiments proved that the model and algorithm proposed in this paper is effective and reliable.The specific innovations are as follows:(1) According to the time characteristics of microblog platform, this paper improves online LDA model and put forward MBO-LDA model, the model is used in modeling online microblog text stream and supervising topic mining algorithm based on comments.(2) This paper put forward Multimedia Microblog Text Description Model with two dimensions of content and emotion. Text is organized into abstract topic tags and specific keywords in each dimension to describe topic of multimedia microblog. This paper also designs a topic mining algorithm to organize text of comments into Multimedia Microblog Text Description Model.