I wonder how easy it would be to make an Irish text-to-speech program (or, equivalently, a program that takes regular orthography and spits out a phonemic transcription, e.g. in IPA or something similar)... that is, how regular the written-to-spoken correspondences are.
I mean, I know that it's not one sound = one symbol, but I wonder how far you could get by taking vowel digraphs/trigraphs, marking them with a triplet of (pronunciation, broad/slender beginning, broad/slender ending), and using that to infer the pronunciation of surrounding consonants.