Semantic-Based Matching and Personalization in FWEB, a Publish/Subscribe-Based Web Infrastructure

The web is a vast graph built of hundreds of millions of web pages and over a billion links. Directly or indirectly, each of these links has been written by hand, and, despite the amount of duplication among links, is the result of an enormous effort by web authors.

One has to ask if it is possible that some of this labour can be automated. That is, can we automate some of the effort required to create and maintain links between pages? In recent work, we described FWEB, a system capable of automating link creation using publish/subscribe communication among a peer-to-peer network of web servers. This allowed web servers to match information about link requirements and page content in circumstances where we specify an anchor in terms of what content we want to link to, rather than a specific URL. When such a match is successful, a link between the pages is automatically created.

However, this system relied on simple keyword-based descriptions, and has several drawbacks, verified by experiment. In this paper, we show how the use of shared ontologies can improve the process of matching the content requirements for links and the descriptions of web pages. We report on our experience of using FWEB and, in addition, show how the capabilities of the FWEB architecture can be extended to include link personalization and explicit backlinks.