Research On Supplement And Quantification Of Network Topics In Forum

There are many topics in network forum, we can comprehend current socialopinions through these topics. We also can monitor them and guide them positively.For the corresponding monitoring system, the main research contents include the topicidentification, topic detection and tracking. Topic identification deals with the networktext set, which contains topic categories and classifies every text in the set intocorresponding category. Topic detection and tracking can monitor, track and controlthe relevant topics.The important and fundamental technologies in existing research are the textclassification and clustering on network topic identification. For a given topic, if usingclassification or clustering method in the documentation set to obtain thecorresponding text collection, regardless of the training sample of learning orclustering number given, or decision category size threshold selection, which is underthe special circumstances of the case in network forum. The result may not be veryaccurate, comprehensive description of the topic. Based on these cases, thedissertation puts forward a network forum topic supplement problem, and realizes atopic supplement method based on keywords. Users can get overall keywords aboutone topic relatively which they are interested in by giving a few words about thetopic.In terms of topics heat calculation, the existing researches use page view, replyquantify, clicks and so on to calculate heat of topics by using weighting treatment. Inthis way, the weights which can be selected will effect great influence on result ofquantization. Choose different weights to calculate heat of the same topic may show asubversive change in the result of calculation. According to this situation, thedissertation advances a calculation method based on users’ activity to calculate topics’heat, which views users’ network behavior as an index in computing network topics’heat. By this method, we can avoid the situation effectively, which is that topics’ heatquantification distorted by a few number of network users who pay great attention tothe topics.