download pdf version

Introduction For many Americans, a request to "transport cargo by lorry" would result in an incredulous stare. In the UK, the term "lorry" means precisely the same thing as "truck" in the US. It is just a matter of semantics, but this minor misunderstanding between English-speaking countries is just one example of the challenges for computer systems trying to compare and correlate data from different information sources that use different semantics. Today Information sources, like databases, may be described using ontologies. An ontology is simply a description for a set of concepts—objects, actions, ideas, etc. An ontol-ogy organizes concepts in a way that represents their real-world relationships in a man-ner that is understood by com-puters. Typically, an ontology is created by a community that shares a common interest—or domain—and needs to describe concepts so that all members of the community understand each concept. The problem occurs when different ontologies are created for the same domain by different communities due to regional, cultural, educatinal, or language differences. Consider, for example the domain for trucks, —motorized wheeled vehicles with a drivers's cab and a rear freight compartment. Creating a world-wide truck database from exist-ing sources would require transforming many different ontologies with different structures and languages. Through continuing research, ATL is developing semantic interoper-ability tools that can examine differ-ent ontologies and match ontological descriptions of concepts (e.g., motorized, cab for the driver and a sepa-wheeled, land vehicle with a rate rear compartment for transporting freight) rather than specific terms such as "truck." ATL's ontology alignment tool, Ontrapro, quickly and automatically discovers semantic alignments between concepts in widely diverse heterogeneous information sources (each described by a unique ontology). It intelligently takes advantage of the many types of descriptive features useful in aligning differ-ent descriptions of common concepts including syntactical, lexical, phonetic, and structural similarities. The goal of ATL's research is to develop "semantic adaptors" that act in much the same way as a human translator: listening to messages from a system with one ontology and transforming them correctly into messages understandable with another ontology. ATL is a leader in semantic alignment research. In concert with its research, ATL has played an active role in the research community in an effort to accelerate mutual progress on this critical challenge. ATL is committed to overcoming the language barriers in a world of different ontologies, data models, and information sources.