Ontology Based Natural Language Discovery of Semantic Web Services in the Geospatial Domain
Yuzhu Yang · NORA - Norwegian Open Research Archives · 2009
Semantic Web Services (SWS) has been an active research topic in the recent years. It aims to achieve automatic Web service discovery, and invocation and composition of services, by adding semantics to the services, which is very significant for accelerating the achievement of Serviceoriented architectures by solving a lot of interoperability issues. Geospatial data on the Web contains more and more important information. Capturing, analyzing and managing such data can help people to achieve various kinds of tasks. Discovery of Semantic Web Services in Geospatial domain is therefore significant. Nowadays the existing semantic Web service discovery frameworks in geospatial domain have either complex interface or require extra knowledge (for instance, the Web Service Modeling Language (WSML) to be able to provide correct queries. This thesis suggests an Ontology Based Natural Language Discovery of Semantic Web Services (ONALDIS). ONALDIS works generally in the geospatial domain, and offers a command line interface which takes natural language questions as input, and gives a WSML query as output. This WSML query can be further processed by an existing SWS discovery framework to discover semantic Web services. ONALDIS achieves the mapping between natural language query and WSML query by analyzing domain ontologies with some rules and generates word sets and WSML queries from the ontologies. In this thesis, semantic Web services and geospatial domain are first introduced. The SWING system and the need for a natural language interface are then discussed. Some existing approaches for natural language interfaces, mainly Question Answering (QA) systems are presented. Further, ONALDIS is introduced as a solution of a natural language interface for semantic Web service discovery in the geospatial domain, with an implementation for the SWING system. Finally, the conclusions are presented and further work is suggested.