The Research And Implementation Of Personalized Push System Of Papers

As a new method for academic publication and scientific communication, the paper published online platform has drawn much attention over few years. However, due to simply set out the information there exist many problems in traditional platform, such as lack of experts, low quality of review and difficult improvement of papers. In the new platform the information related to the research field of experts can be provided actively while that irrespective will be avoided, since the research theme remains unchanged during a long period of time.In this paper, a personalized push system of papers is designed and implemented, which can model the research fields of each expert and dispatch the new published papers accurately based on the principles of text classification. A system to collect paper information is mainly designed to get the papers of experts and new papers on the web-paper library. At the same time we talk about the core module, called the text classification system, we implement the vector space model, Bayesian model to creat the interest model of experts with the improve of the feature extraction method . We analysis the original feature method of mutual Information,then we get this new algorithm of feature method which takes such as frequency,distribution and concentration into account,and it enables the selected feature terms to reach to optimization.Finally comparative experiments show that the Bayesian model is better than the vector space model to express the user's interest and changes. The new algorithm of feature extraction can improve the recall, the accuracy and the value of F1.