In this paper we give a preliminary report on our study of the use of web server traffic logs to improve local search. Web server traffic logs are, typically, private to individual websites and as such – are unavailable to traditional web search engines conducting searches across multiple web sites. However, they can be used to augment search performed by a local search engine, restricted to a single site. Web server traffic logs, which we will refer to as simply logs throughout this paper, contain information on traffic patterns on a web site. By using this information, instead of pure link counts in the computation of PageRank, we can obtain a new local measure of web site importance, based on frequency of visits to a page, rather than simply on the amount of links. In this paper we describe the architecture of a search engine we have built for the Eastern Kentucky University (EKU) website and some preliminary experiments we have conducted with it.
Available at: http://works.bepress.com/dekhtyar/38/