Toward the Union of Databases and Document Management:
                      The Design of DocBase



		           Arijit Sengupta

		Department of Computer Science
		Lindley Hall 215
		Bloomington, IN 47405
		+1 (812) 855-3703 (voice)
		+1 (812) 855-4829 (fax)
		asengupt@cs.indiana.edu (email)


                               Abstract

With the advent of the World Wide Web (WWW) and the increased use of
electronic documents in almost all aspects of computing, the problems
of management of and systematic information retrieval from electronic
documents have become highly pertinent.  Information retrieval (IR)
techniques allow us to retrieve documents based on keywords, but often
these searches are not powerful enough to accurately extract the most
relevant information. Most IR systems are designed for broadening the
scope of search. However, extracting only documents highly related to
the search is often desirable to keep the result set small. We propose
a method for achieving very powerful searches on tagged documents by
using the structural information in the tags as meta-data in
queries. We adopt SGML as our tagging format.  Since HTML is an
application of SGML, and XML is designed as a subset of the SGML
standard, our work is immediately applicable to the current and future
incarnations of the WWW. In this paper, we give an overview of the
methodologies (such as query languages, query interfaces, and query
processing techniques) used in the design of DocBase, our
prototype proof-of-concept document database system. DocBase is a
modular system capable of performing SQL-like queries on native SGML
documents using pluggable indexing and storage-management
applications. Because of the generalized nature of SGML, and the ever
increasing use of structured documents in the corporate world, we
argue that systems like this will be indispensable in the forthcoming
century.