Learning and Application of Differential Grammars

We examine the Differential Grammar , a representat ion designed to discr iminate which of a set of eonfusable al ternat ives is most likely in the context it occurs in. This approach is useful whereever uncer ta inty may exist about the ident i ty of a token or sequence of tokens, including in speech recognition, optical character recognition and machine t ransla t ion. In this paper our appl ica t ion is word processing: we discuss mul t ip le models of confusion which may be used in the identification of confused words, we show how significant contexts may be identified and condensed into Differential Grammars , and we contrast the performance of our implementa t ion with tha t of two commercial g r ammar checkers which purpor t to handle the confused word problem.