Skip to Main Content
In order to share data from disparate date source creating a data transform is an essential step if they have heterogeneities in data representation. Manually creating a transform or finding a transform from a repository of transform is extremely tedious and error-prone. Recent works on data transforms have focused on finding a data transform as a part of the data mapping or schema matching problem. However those works concentrated mostly on structural data mapping and applying a single data transform between schemas. In this paper we describe our approach to data transform composition which reuses existing transforms in a repository for constructing desired transform. The desired transform and existing transform are represented in our transform model as RDF triples. The transform model includes not only semantics of the inputs and outputs, but also the behavior of a transform. Based on ti transform model, the composition problem is treated as a graph search problem. Finally, the overall architecture of a system which automatically discovers a desired transform is described.
Date of Conference: 14-16 May 2008