Presented at: 6th Annual European Semantic Web Conference (ESWC2009)
Research in ontology learning from text has been mainly focused on entity recognition, taxonomy induction and relation extraction. In this work we approach a more challenging research issue, consisting in detecting semantic frames from texts, and using them to encode web ontologies. We exploit a new generation frame detection system, which is able to identify FrameNet frames and their argument boundaries from text parsing. In addition, we enrich frames with the results provided by a super-sense tagger, which extracts and classiﬁes concepts and named entities from texts according to WordNet super-senses. The enrichment results include argument restrictions for the elements of a frame, and domain specializations, based on domain lexical unit detection for the target of a frame. The results are encoded according to the Lexical MetaModel, which allows a complete translation of lexical resources, keeps trace of the learning metadata, and enables custom transformations of enriched frames into modular ontology components, known as content ontology design patterns.
Keywords: Frame Detection, Lexical Ontologies, NLP, Ontology Design, Ontology Learning, Data, Evaluation, Interoperability, Language Technology, Machine Learning, Natural language processing, Natural Language Processing, Ontology (computer science), Ontology (Computer Science), Ontology Engineering, Semantic Web
Resource URI on the dog food server: http://data.semanticweb.org/conference/eswc/2009/paper/143
Explore this resource elsewhere: