“How the World Shapes the Meanings of Our Terms”, in UB’s 6th Annual Postdoc Research Symposium, State University of New York at Buffalo, NY, June 3, 2014.
[ poster ]
This interdisciplinary work in the fields of terminology, ontologies, and natural language processing (NLP) is concerned with natural language definitions in specialized dictionaries and ontologies. It focuses on the modeling of definition contents in view of creating language- and domain-independent tools to help terminologists and ontologists write definitions.
The underlying hypothesis for creating such generic definition-content models is that definitions express characteristics of the types of things to which specialized terms refer. Testing this hypothesis requires using a representation of the types of things that exist in the world, i.e. an ontology. This work uses the categories of the realist upper-level Basic Formal Ontology (BFO). Testing the hypothesis also requires applying these models on corpora of existing definitions to see to what extent the models predict definition contents.
The methodology consists in creating the definition models on the basis of the characteristics of the BFO categories, and using these models as a metalanguage to annotate large-scale multilingual and multi-domain corpora of textual definitions. This postdoctoral research aims at producing the necessary resources for such large-scale corpus analyses: the definition models; definition corpora; computer programs for automatic corpus pre-annotations; an annotation manual for an annotation campaign to create validated training-corpora and to obtain preliminary results.