摘要: Traditional Artificial Intelligence like traditional and modern Linguistics and Computational Linguistics tends to treat language as an isolated phenomenon, and syntax and semantics as separate areas of study. The approach taken to developing grammars, parsers and natural language systems has been more like the dissection of a cadaver than a study of live interactions in a complex ecology: the basic anatomical structure can be discovered and some educated guesses made about the physiological interactions, but there is no hope at all of understanding the functional operation of a system that involves multiple individuals living in a complex environment. In fact, Natural Language research has been even more hampered, as computer scientists ignored the collective wisdom of Linguistics, Psychology and Sociology, and relied on their own intuitions rather than real world data about language in action.This is the context of the 25-year odyssey into Computational Language Learning that has led the present author to an increasingly ecological approach to the study of language acquisition and the engineering of learning systems. Each step along this path has involved the rejection of basic, if often implicit, assumptions of one or more of the cognitive sciences. Indeed, the emergence of Cognitive Science as a coherent field in its own right is itself a precursor of an ecological approach, recognizing that the complex interactions involved in cognition, and in particular language and learning, cross many disciplinary boundaries. At the core of Cognitive Science is the further principle that theories of perception and cognition, and in particular language …
参考文章(0)