Multidimensional Ontologies: Integration of Frame Semantics and Ontological Semantics

TitleMultidimensional Ontologies: Integration of Frame Semantics and Ontological Semantics
Publication TypeConference Paper
Year of Publication2008
AuthorsBarzdi\d{n}š, G, Gr\={u}z\={i}tis, N, Nešpore, G, Saul\={i}te, B, Auzi\d{n}a, I, Lev\={a}ne-Petrova, K\={i}ne
Conference NameProceedings of the {XIII} Euralex International Congress
Conference LocationBarcelona

Today FrameNet–a state-of-the-art implementation of frame semantics-provides one of the best insights into lexical semanticsandtheir interaction with the syntactic structure of the sentence. Themain limitation of the current implementation is the insufficientlevel of formalization of frame descriptions, making it unsuitableforautomatic text annotation without human supervision. Meanwhile,FrameNet usability would greatly benefit from more rigorousformalization and the consequential possibility for automaticannotation. Previous attempts at formalization have focused onenforcing strict ontological control of the semantic types for theframe fillers-despite their insignificant use-due to high ambiguity-inthe actual FrameNet. We propose a different approach relying onrepresentation of FrameNet as a 4D multidimensional ontology thatallows capturing of the precedent knowledge encoded in the manuallyannotated texts, like FrameNet's full-text annotation reports. Thisallows both to re-create FrameNet ontology from semantically annotatedtexts, as well as to use this representation for semantic annotation of new texts. Further extensions of this approach with 5th dimension for anaphora annotat on is discussed as an alternative for the informal semantic type mechanism of FrameNet.