Highly personalized information delivery to mobile clients

摘要

The inherent limitations of mobile devices necessitate information to be delivered to mobile clients to be highly personalized according to their profiles. This information may be coming from a variety of resources like Web servers, company intranets, email servers. A critical issue for such systems is scalability, that is, the performance of the system should be in acceptable limits when the number of users increases dramatically. Another important issue is being able to express highly personalized information in the user profiles which requires a querying power as that of SQL on relational databases. Finally, the results should be customized according to user needs and preferences. Since the queries will be executed on the documents fetched over the Internet, it is natural to expect the documents to be XML documents.

This paper describes an architecture for mobile network operators to deliver highly personalized information from XML resources to mobile clients. To achieve high scalability in this architecture, we index the user profiles rather than the documents because of the excessively large number of profiles expected in the system. In this way all queries that apply to a document at a given time are executed in parallel through a finite state machine (FSM) approach while parsing the document. Furthermore the queries that have the same FSM representation are grouped and only one finite state machine is created for each group which contributes to the excellent performance of the system as demonstrated in the performance evaluation section.

To provide for user friendliness and expressive power, we have developed a graphical user interface that translates the user profiles into XML-QL. XML-QL's querying power and its elaborate CONSTRUCT statement allow the format of the results to be specified. The results to be pushed to the mobile clients are converted to Wireless Markup Language (WML) by the delivery component of the system.