The promise of Semantic Web Services, that of automatic discovery and configuration of semantically described web services, depends on the existence of high quality ontologies that describe the domains of web services as well as their main functionalities. While only few ontologies designed for web service description exist, building such ontologies is time consuming and difficult.

To address this problem, we have built two semi-automatic methods for extracting web service ontologies from textual sources attached to web services (or their underlying implementations). The first method uses extraction patterns applied on the output of a Part Of Speech Tagger, also called surface patterns. The second method relies on deeper linguistic analysis, by employing a dependency checker. This allows writing more complex extraction patterns (called syntactic patterns) and therefore identifying more ontological elements (subsumption hierarchy, meronymy) than with the first method. The talk will describe these extraction methods and their evaluation in two real-life case studies.

Download PowerPoint Presentation (512Kb ZIP file)

Watch the webcast replay >>