Working Group: Classification, Representation and Modeling

This report briefly summarizes the discussions carried out in the working group on classification, representation and modeling of uncertain data. The discussion was divided into two subgroups: the first subgroup studied how different representation and modeling alternatives currently proposed can fit in a bigger picture of theory and technology interaction, while the second subgroup focused on contrasting current system implementations and the reasons behind such diverse class of available prototypes. We summarize the findings of these two groups and the future steps suggested by group members. 1 Theory and Technologies The Big Picture Uncertainty modeling, viewed as an interaction between theory and technologies, is a space of multiple dimensions that cover concepts, theory platforms, technology platforms and application features. More specifically: A Theory Matrix that is the product of a vector of uncertainty concepts and a vector of possible theory platforms represents multiple uncertainty formalisms. Example uncertainty concepts include tuple uncertainty (where the existence of a database record is an uncertain event), and value uncertainty (where values of some of the attributes/features of some entity are uncertain or unknown). Example theory platforms include probability theory and fuzzy logic. The interaction between possible uncertainty formalisms and current technologies (e.g., relational and XML data engines) has produced diverse uncertaintyaware technologies currently proposed by different research groups. For example, introducing uncertainty formalisms based on tuple and value uncertainty and probability theory in relational platform produced probabilistic database engines such as ORION [1] and TRIO [2–4]. Finally, the features implemented Dagstuhl Seminar Proceedings 08421 Uncertainty Management in Information Systems http://drops.dagstuhl.de/opus/volltexte/2009/1941