Second European Conference on Speech Communication and Technology

Genova, Italy
September 24-26, 1991


Computing Linguistic Knowledge for Text-To-Speech Systems With PROSO

Rodolfo Delmonte, Roberto Dolci

University of Venice - Department of Linguistics, Venice, Italy

In this paper we present a computer program PROSO a linguistic rule compiler which applies a number of prosodic rules inserting appropriate markers to any text inputted, on the basis of phonological rules and of syntactic structure, this one computed separately by an RTN parser(see Delmonte & Dolci, 1989) and made visible to PROSO only at a certain structural level, though. The output of this program can then be passed on to a system like VOXPC, the text-to-speech module commercialized by Olivetti, which allows the user to provide explicit phonetic and prosodic information in order to modify the internal coding procedures. VOXPC is a system completely lacking in linguistic knowledge apart from the well-known distinction between content and function words: the quality of the system is very poor both at word level and at sentence and text level.

Full Paper

Bibliographic reference.  Delmonte, Rodolfo / Dolci, Roberto (1991): "Computing linguistic knowledge for text-to-speech systems with PROSO", In EUROSPEECH-1991, 1291-1294.