Information filtering is an important method to alleviate information overload. How to express user's information needs and how to rank them are two main problems in this area. This paper focuses on these two problems. Bookmark service, the solution that is an extension of bookmark function in browser, is used to catch user's information needs. The evaluation of the documents is used to represent user's information needs. Maxine Margin method, a novel ranking algorithm, is designed to increase precision. An experiment proves that it performs better than the traditional methods. A system called Bookmark Service System is designed for implementation. The system construction is given, and its main function is also introduced.