Date of Award

Fall 2007

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Information Systems - (Ph.D.)

Department

Information Systems

First Advisor

Murray Turoff

Second Advisor

Michael J. Chumer

Third Advisor

Jerry Fjermestad

Fourth Advisor

Robert Michael Klashner

Fifth Advisor

Vassilka D. Kirova

Sixth Advisor

Denis L. Blackmore

Abstract

Data Definition Languages (DDLs) have been created and used to represent data in programming languages and in database dictionaries. This representation includes descriptions in the form of data fields and relations in the form of a hierarchy, with the common exception of relational databases where relations are flat. Network computing created an environment that enables relatively easy and inexpensive exchange of data. What followed was the creation of new DDLs claiming better support for automatic data integration. It is uncertain from the literature if any real progress has been made toward achieving an ideal state or limit condition of automatic data integration. This research asserts that difficulties in accomplishing integration are indicative of socio-cultural systems in general and are caused by some measurable attributes common in DDLs. This research’s main contributions are: (1) a theory of data integration requirements to fully support automatic data integration from autonomous heterogeneous data sources; (2) the identification of measurable related abstract attributes (Variety, Tension, and Entropy); (3) the development of tools to measure them. The research uses a multi-theoretic lens to define and articulate these attributes and their measurements. The proposed theory is founded on the Law of Requisite Variety, Information Theory, Complex Adaptive Systems (CAS) theory, Sowa’s Meaning Preservation framework and Zipf distributions of words and meanings. Using the theory, the attributes, and their measures, this research proposes a framework for objectively evaluating the suitability of any data definition language with respect to degrees of automatic data integration.

This research uses thirteen data structures constructed with various DDLs from the 1960's to date. No DDL examined (and therefore no DDL similar to those examined) is designed to satisfy the law of requisite variety. No DDL examined is designed to support CAS evolutionary processes that could result in fully automated integration of heterogeneous data sources. There is no significant difference in measures of Variety, Tension, and Entropy among DDLs investigated in this research. A direction to overcome the common limitations discovered in this research is suggested and tested by proposing GlossoMote, a theoretical mathematically sound description language that satisfies the data integration theory requirements. The DDL, named GlossoMote, is not merely a new syntax, it is a drastic departure from existing DDL constructs. The feasibility of the approach is demonstrated with a small scale experiment and evaluated using the proposed assessment framework and other means. The promising results require additional research to evaluate GlossoMote’s approach commercial use potential.

Share

COinS