| Recent advances in information and communication technologies have given impetus to Web-centric applications such as e-commerce. E-commerce involves business activities that deal with the exchange of values for goods and services on the Internet where all transactions are driven electronically. Access to data distributed over the Web is becoming increasingly difficult, as determining semantically equivalent data remains a major problem. Web data integration aggregates data from diverse data sources into a cohesive, coherent, and global e-commerce data set. However, Web data integration is a complex process that requires unambiguous explanations to elucidate the underlying integration concepts. Unfortunately, most existing efforts involve significant manual input and adopt ad hoc approaches to data integration, ignoring the theoretical foundations of, and the necessary formalisms to specify, the integration process.; This thesis presents a flexible and correctness preserving approach to Web data integration. This approach leverages algebraic signatures with sound integration principles and algorithms to provide a theoretical foundation that facilitates scalable Web data integration. To identify a suitable integration mechanism, a characterization of e-commerce data sets is presented. We provide an algebraic data model that reduces the topological structure of Web data sources to regular expressions and define algebraic operators and functions to manipulate objects in the algebraic model. A common-term vocabulary is used to semantically describe objects from local data sources to eliminate multiple views of objects. We present a high-level architecture of a Web integration system for e-commerce (WISE) and formally specify the basic components of WISE. Formally specifying an integration system for e-commerce transactions provides a clear understanding of the system and also reveals ambiguities, incompleteness, and contradictions in the informal definition, and thus, permits the correctness verification of the integration components. We provide a graph-based integration algorithm that dynamically identifies and analyses correspondence assertions to integrate local data sources. Finally, we show that our integration result is complete, correct, minimal, and understandable. |
