Research On Hot-Topic Detection And Analysis On Internet Public Opinion

In order to strengthen management and monitor to Internet, collection and analysis of public opinion information is a realistic problem solved urgently for the present government departments. The Publiec Opinions Monitoring and Analyzing System (POMAS) aims to automatieally monitor and analyze the huge mount of Internet Pubic Opinions in real time. An important function of POMAS is the hot-topic detection and analysis on Internet public opinions, which can give the government users a quick understanding and mastery of current hot topic on the Internet. Therefore, in the thesis the hot-topic detection and ananysis on Internet public opinion is researched and the mainly work is as follows:Firstly, combined with the characteristics of actual corpus, an algorithm which automatically detects the hot-topics on Internet public opinion is proposed. Based on Single-pass clustering algorithm, the algorithm introduces several strategies to improve the detection effect. The strategies are based on following ideas:above of all, considering the news'characteristic, named entity such as person names, places, organizations and institution names and titles are given higher weight; then, publication time of the report is more important to judge to whether the report belongs to a topic; finally, a topic may include some sub-topics, which is used to impove the detection effection. After analyzing the characteristics of the hot degree of topic, a method is proposed to calculate hot degree. The experiments show that the proposal algorithrm greatly improves the effect of hot-topic detection.Secondly, to mine hot-topic comments from netizens, a kind of mining method on hot-topic comments from netizens is proposed. Based on word segmentation tools, the method firstly revises informal person names which often appear in netizens'commnents. Then, the frequent patterns from the comments are found by frequent patterns mining algorithm. From the frequent patterns the opinions and standpoints of netizens can be obtained and the result can be showed by visual method.Finally, a hot-topic public opinion detectioning and analyzing system is designed and implemented. The whole system consists of four major modules as web crawling, web preprocessing, detection and analysis on hot-topic public opinion and public opinion search. By the system the practical significance of the hot-topic detection and analysis on Internet public opinion is demonstrated.