Theoretical Framework for Defining Validity and Quality in Modeling

The increasing number of Object Oriented databases and knowledge bases raises the need for some criteria of model validity, modeling guidelines, and quality criteria to be used in modeling. Although some researchers have developed some specialized sets of modeling guidelines, we found no general-purpose framework to evaluate these proposals and customize them. In this paper we develop a theoretical framework for characterizing modeling. Modeling is seen as a transformation process. These transformations must be valid, and must increase the quality of the model. The framework is centered around these two issues of validity and quality. 1. KNOWLEDGE MODELING Modeling is a central activity in databases. It consists of defining a set of constructs (relations, classes, constraints) that, together, capture the contents of interest and exhibit model-specific structural quality (normal forms, schema invariants). Software design is also a modeling activity. According to Jocabson et al. [14], “In design we shape the system and find its form (including its architecture) that lives up to all requirements – including nonfunctional requirements and other constraints – made on it.” In other words, design is a modeling activity selecting a form to capture contents. Developing knowledge bases is also driven by the same concerns of capturing contents while exhibiting some structural quality. Overall, regardless of the domain, modeling consists of selecting a form to capture contents. The same contents can be captured in different ways. While the different materializations of the same contents are semantically equivalent, they are not necessarily equivalent with respect to other nonfunctional criteria (readability, reusability, etc.). Quality is an underlying concern in all applications. Traditionally, the term “quality” is used to encompass compliance with nonfunctional requirements, while the terms, “correctness” or “validity” are used to characterize models that capture the right contents. A number of authors have defined sets of 51 Proceedings of the Eighth International Conference on Information Quality (ICIQ-03)