Welsh letter-to-sound rules: rewrite rules and two-level rules compared

In a text-to-speech synthesis system, input words not found in the system's lexicon are passed to letter-to-sound rules, which derive the word's pronunciation. In Welsh, the letter-to-sound rules must be applied in three passes: firstly, to add epenthetic vowels, secondly, to determine stress and vowel location, and thirdly, to perform grapheme-to-phoneme conversion. To begin with, all these letter-to-sound rules were written in the form of context-sensitive rewrite rules, and were evaluated, giving a 96% success rate. The rules for the second pass were then rewritten in the form of two-level rules, using the PCKIMMO software package. The output was identical to that produced by the second block of rewrite rules. The two-level formalism had advantages in simplifying rules. However, there were difficulties due to the need to force the rules to operate in a deterministic fashion. In a practical text-to-speech system, the rewrite rule formalism would be favoured, despite the greater number of rules and their greater clumsiness, since the critical ordering of rewrite rules easily introduces the necessary determinism.