A Component-Based Framework for the Integration and Exploration of XML Sources

Introduction

The past years were characterized by an enormous increase of data available in electronic form, as well as by a proliferation of query languages, data models and data management systems. In such a scenario, traditional approaches to data management are not capable of guaranteeing the suitable level of access transparency to stored data and, at the same time, of preserving the autonomy of local data sources. This situation favored the development of new architectures for data source interoperability conceived to allow users to query pre-existing autonomous data sources to guarantee the maximum possible transparency, efficiency and effectiveness.

Developing modules that handle the reconciliation of involved information sources plays a relevant role in all architectures for data source interoperability. The definition of these modules strongly relies on schema integration, i.e., the construction of a global schema obtained by merging a set of related schemas (Chua, Chiang, & Lim, 2003; dos Santos Mello, Castano, & Heuser, 2002; McBrien & Poulovassilis, 2003). However, when the involved systems are numerous and/or large, schema integration often produces an over complex global schema, which could be not suited to supply a correct and complete description of the available data. In this case, it appears much more adequate the construction of a source hierarchy, representing the involved sources at different abstraction levels. Essentially, this hierarchy can be obtained as follows: initially, the involved schemas are organized into homogeneous groups by means of data clustering algorithms; for each cluster, the corresponding sources are integrated to obtain a global schema representing it and the obtained global schema are in turn grouped to construct second-level clusters, and a representative schema for each of these new clusters is obtained by performing an integration task. This process is iterated until a unique, highly abstract schema representing all the involved XML sources is obtained.

In order to carry out source integration and clustering correctly, the designer must clearly understand the semantics of the involved information sources. One of the most common ways of deriving and representing source semantics consists in detecting interschema properties (Bergamaschi, Castano, & Vincini, 1999; Castano, De Antonellis, & De Capitani di Virmercati, 2001; Palopoli, Saccà, Terracina, & Ursino, 2003; Rahm & Bernstein, 2001) or source constraints. Interschema properties are terminological and structural relationships involving concepts and objects belonging to different sources; examples of interschema properties are synonymies, homonymies and hyponymies. Source constraints are restrictions involving objects belonging to the same or different sources; examples of source constraints are domain constraints, functional dependencies and referential integrity constraints.

The increase in the number of available data sources favored the development of a large variety of possible data formats; in order to uniformly manage them, the adoption of a unified paradigm is compulsory. In this context, the most promising solution has revealed to be XML. Owing to its semistructured nature, XML can be exploited as a unifying formalism to handle the interoperability of information sources characterized by heterogeneous data representation formats. As a matter of fact, XML has become the de facto standard for information exchange. Most of the current information sources are XML-based or can be easily translated into XML.

The considerations outlined above were the premises for the development of the framework proposed in this chapter. Given an input set of XML sources, our framework is mainly designed to: