I want to enable my Microsoft Custom Speech model to recognize designators containing numbers, chars and dashes, something like this: 12-34 A-56 B78
.
The speech model recognizes numbers and characters correctly. Is there a way to train it so it would output the string 12-34 A-56 B78
when i say "twelve thirtyfour a fiftysix b seventyeight"? I need this for a german speech model.
I've already tried to train a model with 10000
randomly generated strings like the one above. I then trained the model using related text.
Thanks in advance
These are very specific format requirements. Unfortunately, it is currently not possible to get results exactly like this from the speech service. I suggest to do some post-processing on the results to format them this way.