Modeling Semistructured Data by Using Graph-Based Constraints

Abstract

The widespread adoption of semistructured and more in particular XML data has opened a new venue for the application of formal methods to the speci.cation of models for de.ning classes of semistructured or XML documents [1,3]. Most current models for semistructured data are too speci.c to allow immediate comparison with each other, and do not easily support incremental model design. For this reason, we introduce the General Semistructured Meta-Model (GSMM), a simple meta-model for semistructured information which can be applied for the translation to a common formalism of the various abstract models proposed in literature; this approach fosters easy a priori comparison and discussion of concrete models’ features, such as allowed sets of values, handling of object identi .ers, relationships representation; moreover, it supports e.ective inter-model translation and design. Unlike other proposals [2], our meta-model does not rely on an extension of the concept of schema; rather, GSMM is based on a generic graph that can be instantiated into a number of concrete models by providing a) values for a restricted set of parameters (labels) and b) some high-level constraints, themselves represented as graphs. Here, constraints take the role of the schema, imposing constraints directly on the structure of semistructured data provides the designer with a powerful tool for controlling the degree of precision of the structure and introducing .exibility at the data representation level.