On the Problem of Semantic Loss in View Integration

Integration is a key issue in conceptual database design. However, the integration process is very complex and error prone and may often cause semantic loss. Semantic loss, in this chapter, is described as a problem that occurs if one or several concept names and/or dependencies describing the meaning of a concept are lost during the view integration process. Semantic loss often occurs because of the way resolution methods are used today since not only similarities, but also differences between the views have to be identified, resolved, and simplified. The high focus on technical and implementation issues that most of the modeling languages adopt today may also cause semantic loss. In this chapter we argue that a modeling language that instead puts focus on concept names and dependencies should be applied. We also argue and propose alternative resolution and simplification techniques for name conflicts and inter-schema properties that instead of getting rid of concept names and dependencies keep these as long as possible in the integration process. Applying the proposed resolution and simplification techniques might counter an impoverishment of the language used in the views and/or schema and prevent semantic loss. The proposed resolution techniques might even contribute to a semantically richer global conceptual database schema.