An Automatic Text-to-Speech Prosodic Translator for the Synthesis of Italian

A computer programme for automatic text to speech translation of the prosodic features of standard Italian is presented. The programme produces a complete phonemic transcription, exception made for voice/unvoiced discrimination in the pairs /s,z/ and /ts,dz/, as well as for open/closed pairs in vowels /e,o/. The Prosodic Translator (hence PT) assigns a series of 9 stress markers corresponding to primary stress /1,6,7/; secondary stress marker /2/; unstressed syllable /3/; semiconsonant /4/; strengthened long consonants in false diphthongs /*/; subphonemic variations at word boundaries produce destressing, demoting, or schwa-like effects with synaleopha, which are accounted for by previous figures plus /9,5/; elision by assimilation is marked /0/. Breath groups, pauses and pseudopauses are marked by /+, ++, -, --/. Sentence and text boundaries are marked by /#, §/. Carets /<, >/ mark potentially prominent syllables carrying the burden of the intonation pattern; /</ indicates a demotion of prominence. An alphabetic symbol /A,B/ appears at the beginning of the intonation string indicating interrogative or exclamative basical intonative pattern. Machine time required to process 1,000 words on a CDC 7600 has been surprisingly low: 0.250 cp seconds execution time, the programme being written in FORTRAN IV.