Research And Applications On Filtering Technology Based On User-Customization

With the rapid expansion and explosive growth of the Web information, it is more and more difficult to acquire valuable data efficiently. To solve these problems, a lot of information filtering system came into being. However, the current filtration systems are for a fixed theme, the theme does not meet the requirements of individual filters. To address this problem, a text filtering system based on user customization is raised. The main works of this paper are presented as follows:1) In this paper, the technology of key word expansion is proposed with which the problem has been solved that the filtering model trained in the late period cannot sieve the data from database fast and precisely as a result of paucity of key words in the initial topic model, when topic set by users. The method exploits term expansion based on local analysis and obtains the words according to the value of TF-IDF and the frequency in documents as the expanded words to the topic.2) In this paper, it is proposed that text filtering based on two-level key word matching. The text meeting the requirement of the model can be achieved fast and precisely. The model is composed of the key words of the topic the user choose and the terms expanded before.3) A semi-supervised text classification based on small samples is proposed in this paper. It solve the problem that the variants of words in the text cannot be recognized.4) Combining the three methods above, a text filtering system based on user customization is built. The data source of the system are data from weibo, blogs, news and forum. Based on different topics users set, the appropriate source data.The system has been proved in theory and practice, to meet users to quickly customize filtration topic model. In the process of filtering, it is no need to annotate the text manually, which meets the requirement of the independent learning largely.