Ontology-enablement of a system for semantic annotation of digital documents.
William J. Black, Simon Jowett, Thomas Mavroudakis, John McNaught, Babis Theodoulidis, Argyrios Vasilakopoulos, Gian Piero Zarri, Kalliopi Zervanou · 2004
We describe the recent enhancement of the CAFETIERE formalism (Conceptual Annotation of Facts, Events, Terms, Individual Entities and RElations) with the ability to link natural language words and phrases in textual documents with instances and classes from a language-enabled ontology. The language-enabled ontology is one with an index from one or more natural language expressions to each concept (as in WordNet). In an information extraction application. the index, ontology and instance repository are consulted in place of the usual gazetteer prior to the application of the context-sensitive phrase structure rules of the CAFETIERE formalism. Information from the ontology and its instances is cached so that rules can be constrained by properties of objects and can in turn build representations using those properties. We describe the notational extensions to CAFETIERE and give examples of the extraction of event instances in the analysis of texts relative to a specific application ontology. Relevant background is given on the architecture and common annotation scheme of the Parmenides system (FP5 project), in the context of which this work has been done.