Research And Implementation Of Network Public Opinion Topic Detection And Tracking System

More and more people publish and access information, participate in the discussionand express their views through the web today, due to the process of the science andtechnology, so the network public opinion become into the important source ofinformation and reference to the government and enterprise. A hot topic in the networkpublic opinion reflects the winds of public opinion, it is a great significance for thebuilding of a stable and harmonious society to discover the network of public opinioncrisis in time, and to take appropriate measures to control and guide the development ofthe hot topics. However, because the network is large and complex, the number ofInternet users is in sharp increase and their active behavior brings in massive amountsof information, so it is difficult to discover the hot topics.Therefore Topic Detection and Tracking gets more and more attention, it is basedon the text learning. As you know, human’s language is complex and has its logical, andthe computer can not understand, When the text change into the form of a computercapable of handling, it will lost a lot of information. Even after the transformation ofsome simple information, it brings in new problem such as high-dimensional and sparse.So there is capacity ceiling for the effect of the topic detection and tracking. Every stepsof the whole process will have a significant impact on the result of the TDT, such as thequality of the text pre-processing, the selection of the feature extraction algorithm, thechoices and improvements on the text clustering and classification algorithms, and soon.In this thesis, in the situation of each TDT technology now in use has their ownstrengths, and each suites to a different environment, so we here consider and choosefrom a variety of different algorithm to compare the results of topic detection andtracking, then determine the best algorithm. At last, we design the network publicopinion topic detection and tracking system. The system provides the hot topics list forusers. The user can search for stations within a hot topic, select a hot topic in the list,track the topic, and view topic clustering distribution, the topic development trends andrelated information. In addition, the user also can configure and choose their own combination of optimal algorithm or algorithms flexible. This system has the followingfeatures:(1) Integration. The network public opinion topic detection and tracking systemprovide users with a full range of topics related information, including the topic name,topic description, core description of the event, related documentation, etc.; it provideusers with a variety of clustering classification strategy algorithm; topics related to allkinds of information while providing visual icon and report generation.(2) Interaction. Network public opinion topic detection and tracking system allowsthe user to take the initiative to search for hot topics; network public opinion topicdetection and tracking system allows users to freely configure clustering classificationthe policy algorithm or combination of algorithms, in order to meet the conditions of theactual situation, and achieve optimization of topic detection and tracking.