We describe an automatic system for generating phonotypical prosodic items from orthographic texts and their syntactic-prosodic underlying structures. The method we use involves a morphophonological lexicon and a phonological component that relies on the notion of multifold pronunciation group. This allows us to account for pronunciation variants by means of one single phonotypical transcription of a speech item. Corpora of such items can be used as input of an automatic alignment system for the purpose of automatic labelling of recorded speech corpora. The method can also be applied for the purpose of phonological knowledge representation of automatic speech recognition systems. Keywords: Speech Recognition, Automatic alignment, Prosody, Phonology, Phonetic
Bibliographic reference. Tiboni, J. / Perennon, G. (1991): "Phonotypical transcription through the GEPH expert system", In EUROSPEECH-1991, 767-770.