Dynamically exploiting available metadata for browsing and information retrieval

Abstract

Systems trying to help users deal with information overload need to be able to support the user in an information-centric manner, and need to support portions of the information which are structured -like creation dates -while at the same time allowing for irregularity and evolution of the data schemas which are not allowed in databases, i.e. they need to support semistructured repositories. Thus, to solve information overload there is a need for a system to support finding information in these semistructured repositories. Since users rarely know the exact query to ask a system when searching for information the first time, they will want to take the results from their first query and use it to successively improve their queries to "home in" in on the needed information. This research investigates a framework, system, and user interface for supporting effective information retrieval and browsing in such repositories. This system is general purpose, involving no hard-coded assumptions about the structure of the repository. An understanding of the end-user's search process is used to provide effective "next steps" to the users leading them towards the information they are seeking.