In order to translate phonological phenomena into speech-decoding systems, we have introduced a phonological model. It's based on contextual phonological groups (cpg's) and multi-pronunciation groups (mpg's). An interesting feature, with such a model, consists in allowing an automatic adaptation of the phonological component to suit a transcription of a corpus read aloud. Thus, the phonology involved can be tailed to a particular speaker or group of speakers. The purpose of this paper is both to describe the bases of the phonological model, and to present the system intended for automatic adaptation of the phonological component

Keywords: Speech Recognition - Phonology -Automatic alignment - Learning