Reading Behavior Based User Interest Model And Its Application In Recommender System

While the scale of data on Internet is growing rapidly,users need more efficient methods to retrieve information.One of widely used methods is recommend system,which is able to filter information based on user interests.User interest model is a core component of recommend system,it is aimed to describe user interests precisely,and to evaluate the degree of user interest in certain document,in order to help decide whether the document is recommended or not.There are three challenges for developing a user interest model,i.e.user data acquisition,user profile representation and description of user interest evolution.In this paper,we propose a user interest model for reading recommend system.We collect user data in both explicit and implicit ways,then create user profiles based on interest tags.After extracting keywords from articles,by analysing the semantic relevance between keywords and interest tags,we are able to calculate the degree of user interest in articles and decide whether the articles should be recommended.We also implement a mechanism to dynamically update user profiles through analysing their reading behaviors,hence reflecting the change of user interests.The proposed user interests model is integrated into a reading recommender system,articles from Internet are recommended to users.The results of experiment validate the the effectiveness of our model.