17 - Non-monotonicity in Linguistics

Abstract

The origins of nonmonotonic logics in AI share a motive with a research tradition that attempts to see linguistic meaning as a systematic source of insights into thought and reasoning. More recent linguistic work attempts to link this tradition to formal logic. Because the projects of formalizing basic areas of common sense reasoning and of providing logical tools for interpreting natural language are so closely related, this is perhaps the most obvious application for nonmonotonic logics in linguistic theory, and there is a growing body of work in linguistic semantics that makes use of nonmonotonic logic. But the relations between nonmonotonic logic and linguistic theory are much more pervasive; nonmonotonicity can be recognized in many areas of linguistic research that have little or nothing to do either with meaning or common sense. Every area of linguistics encounters generalizations that have exceptions. As long as linguistic theory lacks direct a way to represent such generalizations, they can't be expressed in the theories that explain them. In practice, this often means that the generalizations are saved by ad hoc maneuvers. Most of the available applications of nonmonotonicity in linguistics have been developed by computer scientists, or by linguists with computational interests.