Text-to-speech conversion on a personal computer

A system that uses a set of units derived from Arabic syllables to synthesize unlimited-vocabulary speech on PCs is presented. Text-transcription rules operate on input orthography to generate an abstract representation of phonemes and syllables that form synthesis units and speech. General aspects of the problem, such as the synthesis method and the generation of synthesis units, are discussed. The rules and algorithms of speech synthesis and the current PC-based realization of these algorithms are presented. System evaluation results show an average understandability score of 93%. The system requires about 20 s to compose and replay 1 s of speech when using the PC XT and about one third of this figure on the PC AT.<>