I've read some papers about data/information integration. Most of them focuses on the matching operation in the integrating procedure. In my impression, they didn't put much on the merging of different sources. Maybe it lies on some reasons. First, many applications of data integration do not require an integrating view of different sources, like E-Commence, which pays more attention to the communication among peers. Second, in the data integrating system, the matching part seemly is more important than the merging part, because the matching is the fundamental operation in the data integration. Merging should be done based on the result of matching. 

Let's face the truth. We must have a component in the potential system to do the matching work. "A survey of approaches to automatic schema matching" and "Generic Schema Matching with Cupid" are two good references for the work on the area of Matching. When a simple implementation of our algorithm is concerned, we may borrow ideas from others' work. According to Dr.Candan, our focus is on the integrating part: how to integrate ontologies from different sources into one globe view.