Research On Key Technology Of Distributed Public Opinion System Based On CEP

With the rapid development of the Internet, a lot of new information on the Internet is occuring for people to read and learn. In order to grasp these new information users can access a variety of portals or search engines. But the information is not comprehensive enough on portals, and search engines is not timeliness due to the high cost of making index of the information. The public opinion system is made to solve these problems. Public opinion system is a kind of real-time systems to download the latest online news, which has more channels of information broader than portals and more timely than search engines.As the application scenario is changing, the between public opinion system and the user should also be adjusted, the interaction used in the traditional search engine is not suit for the public opinion system. As the public opinion system always has data in realtime, it always been difficult to design a good feedback interactive process. As a result, it is hard to determine the user’s search intention. After the analysis of the problem and consideration of human to human communication, the paper puts forward the concept of consensus understanding. Based on this concept and CEP technology, we designed the distributed public opinion system.Meanwhile, the traditional public opinion system always be designed only to work in the well cared server cluster. As a result, using the system will cost a large amount of computing resources and bandwidth resources. Nowadays the high cost is the major factor which limits the development of the public opinion system. As a result, only the government and enterprises are able to build the fully functional public opinion platform. To solve this problem, the system we designed could use the high-dispersion network resources to work, which greatly reduces the cost of the system deployment and operation. In order to control the stability of this distributed system, the paper made some research of the corresponding problems. For example, in order to solve the operational performance of the distributed systems, this paper propose a management function of "Token". To solve the operational stability of the distributed system, this paper proposes a "Stable-Core" model. At last, this paper made some experiments and analysis to evaluate our approach. The test results showed that the prototype system has a good performance on user intent detection capability and operational performance.