Research Of Hot Topic Discovery Strategy On Microblogging Platforms

In the past a few years, microblogging, has gradually become familiar with people. The microblogging platforms have been developing beyond expectations, and the services they provide are also growing fast. At present, the strategies taken in the industry field for hot topic discovery are comparative rough, and the results of satisfying the users'requirements are not good enough.This article starts with two aspects:latest studies upon microblogging phenomenon and current solutions on hot topic discovery. We made a thorough investigation on the past researches in the microblogging field. After careful analysis on this problem, we choose semantic analysis as our research foundation. Together with the time sequence and text characteristic on the platform we purposed a new iterative model for semantic analysis and hot topic statistic:TopicRank. This model calculates the influence of the topic in continuous time by time slicing and keyword set concepts, and predicts the trends of the topics' influence in the foreseeable future.This article also takes the social network structure in microblogging platform into consideration. Based on the study of the users'profiles and the follow relation between them, we brought in a user-oriented topic ranking method:TopicRank-U. After combination with TopicRank, we purposed a modularized system design plan, and proved its availability and effectiveness through a series of tests and experiments.The strategy proposed in this article introduces obvious improvement to traditional hot topic discovery method. During the design phase of the algorithm, we considered a lot on reducing manual work and raise efficiency, thus this strategy would be of certain practical value.