Abstract:
One of the current goals of the Pangloss knowledge-based MT project is the construction of a large (35000-node) ontology and lexicon. The upper region of the ontology (called the Ontology Base) is a 400-node synthesis of the PENMAN Upper Model and ONTOS. The rest is a network of commonly-encountered objects, processes, qualities, relations, etc., found in the application domain. Manually constructing a network of this scale is time-consuming, so we must turn to automatic methods where feasible. The first stage of our ontology building consists of taxonomizing tens of thousands of English word senses, and subordinating them to the Ontology Base.