hi everybody
When a single word , has 2 or more pronunciations in the lexicon , than which one is placed during the alignment ?
By alignment I mean the "tagging" or labeling the phoneme , or tri-phone per each frame ,which than is used for the NN supervised training.
So if a word has multiply pronunciation , does it splits the alignment to multiply "labels" or chose the most likely pronunciation and use it to label the frames ?
thanks