Stochastic Grammars: Theory and Applications

摘要

An investigation of stochastic grammars and stochastic languages is presented.Their use as a formal model to describe statistical aspects of sentences in natural languages is discussed. Weakly restricted stochastic context free grammars are introduced and compared with the classical model of stochastic grammars. It is argued that a restricted stochastic grammar is a better model for ordering the structures of ambiguous sentences than the usual (unrestricted) stochastic grammars. The notion of consistency for these grammars is defined. Problems concerning parameters inference and stochastic parsing are stated. Several problems concerning inference of stochastic grammars and grammar classification are posed.