Filtration System Based On Mixed-mode Text

When the society has entered the information times, facing the abundant on-line textinformation, how to help the user find the information that is interested is a very important task..Text filtering can help the user to obtain the texts which they are interested in and can realizepersonalized information service. Therefore, text filtering is of great value and used widely.There are two kinds of text filtering: Content-based Filtering and CollaborativeFiltering.Content-based Filtering mainly adopts some technologies such as natural languageprocessing,artificial intelligence and probability statistic to analyze text content., then calculatedegree of similarity between content vector and user profiles vector and select high correlativetext to registered users. However, it is difficult for content-based filtering to distinguish thequality of the filtered results of the same topic. So, it can not find the new information that theusers are interested in. Collaborative filtering mainly makes use of users' opinions who havesimilar interest to predict and recommend. Now it has been used in personalized recommendationsystem. But with the system scale enlarging, its efficiency gradually decline and some problemssuch as Sparsity and early rater will appear.The paper first describes the two kinds of the filtering methods. Then we study them separately.We put the focus on the study mechanism of the Content-based filtering. We treat the onlineinformation filtering as a reinforcement learning process. A method is then presented thatacquires reinforcement signals automatically by estimating user's implicit feedback. Usingreinforcement learning, a model for adaption information filtering improve the performance ininformation quality and adaptation speed. About collaborative filtering, we propose aprobabilistic user-item relevance model which can effectively improve the precision of therecommendation and settle the existed the problems. By making the experiment, it shows that themodel can improve the performance of the collaborative filtering system. Lastly, we combine thetwo kinds of the filtering methods and propose a hybrid model for text filtering.Then ,we make experiments based on the content-based text filtering,collaborative filteringand the hybrid model filtering method separately. We compare the group results obtained fromthe three kinds of methods. It show that the hybrid model can improve the precision of thefiltering system and has better performance than the other methods. In the end, we point out theshortage and the problems existing the study and the paper. and the goal that we will reach in thefuture.