Swarm intelligent surfing in the Web

Abstract

Traditional ranking models used in Web search engines rely on a static snapshot of the Web graph, basically the link structure of the Web documents. However, visitors' browsing activities indicate the importance of a document. In the traditional static models, the information on document importance conveyed by interactive browsing is neglected. The nowadays Web server/surfer model lacks the ability to take advantage of user interaction for document ranking. We enhance the ordinary Web server/surfer model with a mechanism inspired by swarm intelligence to make it possible for the Web servers to interact with Web surfers and thus obtain a proper local ranking of Web documents. The proof-of-concept implementation of our idea demonstrates the potential of our model. The mechanism can be used directly in deployed Web servers which enable on-the-fly creation of rankings for Web documents local to a Web site. The local rankings can also be used as input for the generation of global Web rankings in a decentralized way. © Springer-Verlag Berlin Heidelberg 2003.