Cager: a framework for cross-page search

Abstract

Existing search engines have page as the unit of information of retrieval. They typically return a ranked list of pages, each being a search result containing the query keywords. This within-one-page constraint disallows utilization of relationship information that is often available and greatly beneficial. To utilize relationship information and improve search precision, we explore cross-page search, where each answer is a logical page consisting of multiple closely related pages that collectively contain the query keywords. We have implemented a prototype Cager, providing cross-page search and visualization over real dataset.