Learning Stochastic Categorial Grammars
Miles Osborne, Ted Briscoe · 1997
Stochastic categorial grammars (SCGs) are introduced as a more appropriate formalism for statistical language learners to estimate than stochastic context free grammars. As a vehicle for demonstrating SCG estimation, we show, in terms of crossing rates and in coverage, that when training material is limited, SCG estimation using the MinimumDescription Length Principle is preferable to SCG estimation using an indifferent prior. 1 Introduction Stochastic context free grammars (SCFGs), which are standard context free grammars extended with a probabilistic interpretation of the generation of strings, have been shown to model some sources with hidden branching processes more efficiently than stochastic regular grammars [12]. Furthermore, SCFGs can be automatically estimated using the Inside-Outside algorithm, which is guaranteed to produce a SCFG that is (locally) optimal [1]. Hence, SCFGs appear to be suitable formalisms for the estimation of wide-covering grammars, capable of being used...