A Frame of Personalized Information Filtering System Based on XML

To cope with the problem of information overload, various information filtering system have been proposed. However, most of them represent documents and user's interests as bag of words, and thus the intrinsic structures and semantic information in them are neglected. The information filtering mainly depended on the matching of simple key word or bag of words. In this paper, a multi-agent personalized information filtering system based on XML has been developed, under which heterogeneous, distributed data on the Internet are integrated to XML data, and the similarity of documents to user's interests is measured using an extended vector space model by which documents and user's interests are represented as structural vectors taking both content and structure into account. The general architecture of the system is presented and several key technologies for the construction of such a system are also described.

1. Introduction

With the rapid development of the Internet technology, information on the Internet has been growing explosively. Locating relevant information on the Internet is time-consuming and expensive. To help people overcome the difficulty of information overload, various personalized information filtering system have emerged [3], [4], [7], [11], [12]. However, most of them represent documents as a set of key words of mutual independent, and thus the intrinsic structures of the documents and the semantic information embodied in them are neglected. The information filtering mainly depended on the matching of simple key word or bag of words. That is to say, the traditional information filtering system can not very well describe and depict document information, user interests and the similarity between them, which results in the users being surrounded and puzzled by vast irrelevant information and failing to find their interested information.