Tagging with Queries: How and Why?

Abstract

Web search queries capture the information need of search engine users. Search engines store these queries in their logs and analyze them to guide their search results. In this work, we argue that not only a search engine can benefit from data stored in these logs, but also the web users. We first show how clickthrough logs can be collected in a distributed fashion using the http referer field in web server access logs. We then perform a set of experiments to study the information value of search engine queries when treated as "tags" or "labels" for the web pages that both appear as a result and the user actually clicks on. We ask how much extra information these query tags provide for web pages by comparing them to tags from the del.icio.us bookmarking site and to the pagetext. We find that query tags can provide substantially many (on average 250 tags per URL), new tags (on average 125 tags per URL are not present in the pagetext) for a large fraction of the Web.

Available Versions of this Item

• Tagging with Queries: How and Why? (deposited 19 Nov 2008 09:34) [Currently Displayed]
  • Tagging with Queries: How and Why? (deposited 17 Feb 2009 08:11)