This statement of interest presents a brief rationale and description of issues for using ontology alignment as a key step in dynamically chaining together a sequence of web services. Communicating author: Dru McCandless, The MITRE Corporation, 1155 Academy Park Loop, Colorado Springs, Colorado 80910.
As in much of the world, the Department of Defense (DoD) has seen an explosion in the growth of web services. But integration of these disparate information sources to answer complex questions remains a challenge. Many information integration tasks are unforeseen at the time the services are constructed, and are therefore difficult to perform “on the fly”. This typically involves searches among various web service definitions and deciding how best to arrange and call them in an ad-hoc manner. A better method of assembling a dynamic service chain is needed.
Using semantic web technology to semi-automatically create a service chain
is an active area of research [
We have developed a different approach for dynamic web service assembly
that takes advantage of the formal structure inherent in web services that are defined
by WSDL documents. This is based on our past efforts using ontology alignment to
integrate different sources of information [
We are sometimes asked to justify converting schemas into ontologies to do alignment when there are schema alignment tools available. Our response is that by converting to an ontology, it enables us to apply the power of the underlying logic model to make better decisions – an example being the case where there are two schemas with the word „mustang‟, but one refers to the car and the other the horse. A purely linguistic aligner will almost always align these – and usually the schemas we work with are small enough such that a structural analysis doesn‟t have enough information to make a better decision. But by using ontologies, it is fairly easy for a person to add some additional taxonomy information above each of the „mustang‟ classes, by asserting for example that one mustang is a subclass of vehicle and the other is a subclass of animal, and that the two classes are disjoint. This should enable an ontology matcher to reach the correct conclusion. In addition, of course, using an ontology enables one to perform automated consistency checking on it – something that is not easy to do with a schema. As alignments become complicated, and the ontologies involved become large and complex, consistency checking becomes increasingly valuable.
In practice, when applying ontology alignment to real-world services a number of difficulties are encountered, with missing and false alignments being the most frequent. Some of the areas about which we would like to engage alignment researchers are: an increased emphasis on meta-properties, such as disjointness, to help with alignment decisions, and techniques for including domain information (such as mid-level ontologies or controlled vocabularies) to improve performance.