Department of Software Technology
Vienna University of Technology
Automatic Classification of Semantic Concepts in View Specifications
The design of large database systems often is done by a large number of analysts with different perspectives on the problem domain. That is why the integration of multiple views in database design is a task of tremendous importance. In this paper we report on our experience in using a view comparison tool based on neural network technology. The tool automatically extracts semantic concepts from different view specifications and then transforms them into a vector representation understandable by a neural network. The network is trained and thus performs the job of clustering similar concepts. The output of our tool is a `first guess' which concepts in views may overlap or which concepts do not overlap at all. The two main contributions of our tool thus are first, that the designer is relieved from the burden of manually comparing each semantic concept of the different view specifications, and second, that human interference into the view comparison process is minimized to the specification of views and the interpretation of concept clusters.
Up
Comments: rauber@ifs.tuwien.ac.at