The Inside-Outside Algorithm: Further Experiments on Three and Four Letter Word Spelling and Implementation on an Array Processor.

This paper describes an application of the Inside-Outside algorithm to learning rules for the spelling of English words. The initial experiments examine the effect of using prior knowledge and further experiments compare stochastic regular grammars with stochastic context-free grammars. The problems of using the Inside-Outside algorithm are discussed and an implementation on an Array Processor is described. Keywords: Great Britain. (fr)