In this chapter we introduce a knowledge engineering methodology to adapt existing portions of FrameNet to new or specialized domains. Firstly, frame occurrences are detected in domain texts by a FrameNet-based statistical analyzer. Secondly, frame arguments are assigned additional semantic types by using a supersense tagging tool. Thirdly, the resulting instances are statistically filtered in order to select the most relevant ones for the specific domain. Finally, we represent the newly created frames as OWL2 ontologies. We exploit state-of-the-art Natural Language Processing technology for frame detection and super-sense tagging. The formal semantics behind OWL2 is used overall to back the learning process: the semantics of frames is discussed, and choices are made to maintain the best from the two worlds of lexical and formal semantics, also exploiting the Linguistic Meta Model as a bridge. The proposed methodology is aimed at mostly automatizing the domain adaptation process performed by a domain expert. We retain a human intervention step for final quality assessment of new frames before their inclusion in the specialized domain ontology resulting from the process.
Coppola B, G.A. (2014). Learning Domain Ontologies by Corpus-Driven FrameNet Specialization. Amsterdam : IOS Press.
Learning Domain Ontologies by Corpus-Driven FrameNet Specialization
GANGEMI, ALDO
Membro del Collaboration Group
2014
Abstract
In this chapter we introduce a knowledge engineering methodology to adapt existing portions of FrameNet to new or specialized domains. Firstly, frame occurrences are detected in domain texts by a FrameNet-based statistical analyzer. Secondly, frame arguments are assigned additional semantic types by using a supersense tagging tool. Thirdly, the resulting instances are statistically filtered in order to select the most relevant ones for the specific domain. Finally, we represent the newly created frames as OWL2 ontologies. We exploit state-of-the-art Natural Language Processing technology for frame detection and super-sense tagging. The formal semantics behind OWL2 is used overall to back the learning process: the semantics of frames is discussed, and choices are made to maintain the best from the two worlds of lexical and formal semantics, also exploiting the Linguistic Meta Model as a bridge. The proposed methodology is aimed at mostly automatizing the domain adaptation process performed by a domain expert. We retain a human intervention step for final quality assessment of new frames before their inclusion in the specialized domain ontology resulting from the process.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.