| Current topic opinion analysis systems are related with a single field or the whole internet. The research area is too large, as a result that the depth of the research is not deep enough. The purpose of this issue is to design a new public opinion analysis system, which can provide the detailed analysis on a special event. The analysis result of this system can help the government or some large companies to make decision about some events.The research method of this subject is researching the related technologies at the beginning, then listing the deficiencies of each technology and finding solutions. In this issue, we focus on the topic web crawler, entity recognize technology, document cluster technology and social relationship analysis technology. At first, with a lot of research on related technologies, we provide a distributed topic web crawler based on site search, which is effective and easy to use. What’s more, there is no need to set large number of links for this crawler, and the traditional task management based on links is absolutely abandon in this crawler system. Secondly, this issue introduces a Named entity recognition algorithm based on CRF model, in which we add Series of methods to check the analysis result to improve the accuracy. Thirdly, this issue provides a cluster algorithm based on K-MEANS and LDA which can improve the effect of the process of clustering. What’s more, this algorithm include K-MEANS cluster algorithm which is improved. This Cluster algorithm is more effective than simple K-MEANS cluster algorithm. At last, this paper provides a social entity relationship analysis algorithm, which is based on support probability and belief probability.This paper proves the public opinion analysis system for topic event is useful by designing and implementing the system. This system can point out the public opinion on some special events, and provide advice to some organizations. What’s more, with a lot of research on technology related, this paper provides many good algorithms which are useful for public opinion analysis field. |