Semantic Web Service Usage Tasks in WSMO - Enabling Semantic Web Services: The Web Service Modeling Ontology - page 114

9.2.2 Web Service Discovery
We now discuss in more detail how Web service discovery is addressed within
the WSMO framework. We outline several different levels that require different
efforts for annotation and description of both goals and services and deliver
results of different accuracy. Each approach addresses a different level of ab-
straction in Web service descriptions. The various techniques taken together
help to create a workable solution to the problem of Web service discovery
which addresses practical requirements and is based on realistic assumptions.
Web Services at Various Levels of Abstraction
The descriptions of Web services, as published by service providers, and the
goals of service requesters need to be compared with one another during the
discovery process in order to identify matching Web services and goals. In
essence, matching in the discovery process is about finding common elements
in these descriptions. Depending at the level of detail which we consider these
entities, we end up with models of Web services at varying levels of abstraction:
At the most fine-grained level, we can consider services as concrete state
transitions ( s, s )fromastate s to another state s .Thestates s and s deter-
mine precisely how the world is before and after the service has been provided.
On a more abstract level, we can ignore the detailed structure of services (i.e.
the respective states), understand them purely as abstract objects, and char-
acterize their specific properties. In terms of ontologies, we would then con-
sider services as instances and Web services as concepts. At the most abstract
level, we can abstract even further in the description of abstract services by
ignoring the description of the possible elements of the set. At this level, we
would simply use a term or keyword for describing the abstract service and
neglect any information about the fine-grained structure. Essentially, we are
considering keyword-based descriptions here.
Each of these levels of abstraction implies a different description of Web
services, ranging from detailed characterizations of possible state transitions,
less detailed descriptions using (complex) concepts in an ontology, to sim-
ple unstructured keywords. Consequently, the achievable accuracy of a result
in the discovery process varies significantly, since more or less structure is
reflected in the descriptions. On the other hand, the ease of providing the
descriptions varies drastically between these levels as well. Whereas simple
keywords are easy to provide, the descriptions of concepts are not hard to
provide but this requires more effort. The provision of detailed state-based
descriptions definitely requires more skill from the people creating the formal
descriptions. The more fine-grained the information is, that the descriptions
reveal, the more complex the algorithms must be in order to deal with these
descriptions.
Therefore, there is an interesting trade-off between the possible achievable
accuracy, the ease of creating the descriptions, and the degree of computa-
tional e ciency of the discovery process.