With the appearance of Semantic Web technologies, it becomes possible to develop novel, sophisticated question answering systems, where ontologies are usually used as the core knowledge component. In the EU-funded project QALL-ME, a domain-specific ontology has been developed and applied for question answering in the domain of tourism. This ontology is encoded in OWL DL and covers several important aspects related to the tourism industry, including tourism destinations (i.e. cities and towns), tourism sites (i.e. accommodation, gastro, attraction, and infrastructure), tourism events (e.g. movie and show) and transportation. To expand domain-specific concepts to general terms which are applicable to a wide range of domains, the QALL-ME ontology was aligned with two upper ontologies - WordNet and SUMO, which are widely used in the NLP field. Due to the changing nature of the ontology throughout the life of the project, semi-automatic means for performing this mapping were developed.
After the ontology was designed, it was used to semantically annotate authentic tourism data and natural language benchmark questions created specifically for the project. The tourism data, which are available in either structured (e.g. database, XML) or unstructured (e.g. HTML) format, were annotated with semantic markups derived from the ontology and converted into the RDF format to instantiate the QALL-ME ontology. The RDF data was stored in a relational database using a RDF storage system (e.g. Jena2, Sesame), and SPARQL, a WC3 recommended RDF language, was used to retrieve specific content from the database. The benchmark questions were tagged with Expected Answer Types (EAT) and Named Entities (NE). The QALL-ME ontology was used as the source of labels for EAT and NE tagging, where EATs were defined as ontology concepts, datatype properties, RDF concepts (xsd:string, xsd:int, xsd:date etc.), or their combination, and NEs as ontology concepts or datatype properties.
Shiyan Ou, Viktor Pekar, Constantin Orasan, Christian Spurk, Matteo Negri (2008). Development and alignment of a domain-specific ontology for question answering. In Proceedings of the 6th Edition of the Language Resources and Evaluation Conference (LREC-08). Marrakech, Morocco, May 28th-30th 2008. [pdf, 342 KB]
If you have any problems, please contact: Luisa Perenthaler (perenthalerfbk.eu)
note: _concepts file name represents the mapping to ontology concepts and _properties file name represents the mapping to the ontology datatype properties.