Document Server@UHasselt >
Research >
Research publications >

Please use this identifier to cite or link to this item: http://hdl.handle.net/1942/16395

Title: Discovering XSD keys from XML data
Authors: Arenas, Marcelo
NEVEN, Frank
UGARTE, Martin
Issue Date: 2013
Publisher: ACM
Citation: Ross, Kenneth A.; Srivastava, Divesh; Papadias, Dimitris (Ed.). Proceedings of the ACM SIGMOD International Conference on Management of Data, p. 61-72
Abstract: A great deal of research into the learning of schemas from XML data has been conducted in recent years to enable the automatic discovery of XML Schemas from XML documents when no schema, or only a low-quality one is available. Unfortunately, and in strong contrast to, for instance, the relational model, the automatic discovery of even the simplest of XML constraints, namely XML keys, has been left largely unexplored in this context. A major obstacle here is the unavailability of a theory on reasoning about XML keys in the presence of XML schemas, which is needed to validate the quality of candidate keys. The present paper embarks on a fundamental study of such a theory and classifies the complexity of several crucial properties concerning XML keys in the presence of an XSD, like, for instance, testing for consistency, boundedness, satisfiability, universality, and equivalence. Of independent interest, novel results are obtained related to cardinality estimation of XPath result sets. A mining algorithm is then developed within the framework of levelwise search. The algorithm leverages known discovery algorithms for functional dependencies in the relational model, but incorporates the above mentioned properties to assess and refine the quality of derived keys. An experimental study on an extensive body of real world XML data evaluating the effectiveness of the proposed algorithm is provided.
URI: http://hdl.handle.net/1942/16395
DOI: 10.1145/2463676.2463705
ISBN: 978-1-4503-2037-5
Category: C1
Type: Proceedings Paper
Appears in Collections: Research publications

Files in This Item:

Description SizeFormat
Main article264.31 kBAdobe PDF

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.