Finding pages on the web that are relevant to some user-defined criteria is a longestablished area of research. Early work on search engines concentrated on the textual content of web pages to find relevant pages, but in recent years, the analysis of information encoded in hyperlinks has been used to vastly improve search engine performance. This paper presents a variation on the use of link analysis for automatically categorizing web pages, by defining a similarity measure and search technique

more »

... based on hyperlinks. This measure is used to categorize hyperlinks themselves, rather than web pages. This allows us to, given a particular hyperlink, to find others like it in terms of its context and referred-to web pages. We have implemented a prototype version of the technique as a browser application. Initial tests appear to show that the method, as implemented in our browser, returns precise, well-focussed results.