Most speech synthesizers have tended to depend on letter-to-sound rules for most words, and resort to a small "exceptions dictionary" of about 5000 words to cover the more serious gaps in the letter-to-sound rules. The Bell Laboratories Text-to-Speech system, TTS, takes a radical dictionary-based approach; dictionary methods (with morphological and analogical extensions) are used for the vast majority of words. Only a fraction of a percent (0.5% of words overall; 0.1% of lowercase words) are left for letter-to-sound rules. Moving to an extreme dictionary-based approach cuts the error rate by at least an order of magnitude. Now that the dictionary is the rule and not the exception, the term "exceptions dictionary" seems somewhat dated.
Cite as: Coker, C.H., Church, K.W., Liberman, M.Y. (1990) Morphology and rhyming: two powerful alternatives to letter-to-sound rules for speech synthesis. Proc. First ESCA Workshop on Speech Synthesis (SSW 1), 83-86
@inproceedings{coker90_ssw,
author={Cecil H. Coker and Kenneth W. Church and Maik Y. Liberman},
title={{Morphology and rhyming: two powerful alternatives to letter-to-sound rules for speech synthesis}},
year=1990,
booktitle={Proc. First ESCA Workshop on Speech Synthesis (SSW 1)},
pages={83--86}
}