Search

Although natural language processing can be deemed as a discipline between applied linguistics and artificial intelligence, theoretical linguistics has played a remarkably minor role in this field of research. One of the goals of this paper is to portray the reasons of the failed symbiosis between linguists' research and that of computer scientists, where probabilistic approaches haven been steadily overshadowing linguistic models. In spite of this discouraging scenario, FunGramKB, a knowledge base particularly designed for natural language understanding systems, serves to illustrate how a language-aware and cognitively-plausible approach to human-like processing can contribute to the development of enhanced knowledge-engineering projects. [ABSTRACT FROM AUTHOR]

This paper describes the methodological aspects of the construction of terminological subontologies within the architecture of FunGramKB. The main hypothesis is that the multilevel model of the Nuclear Ontology in FunGramKB (metaconceptual level, basic level and terminal level) can be connected to a terminological subontology in order to minimize redundancy and maximize information knowledge. One of the features of this subontology is, firstly, its implementation for the creation of a specialized module in the Ontology and, secondly, its focus on deep semantics, in contrast with the vast majority of terminological ontologies, which are based on surface semantics. [ABSTRACT FROM AUTHOR]