Open JerameyATyler opened 4 years ago
This endpoint needs to receive a text transcript and a language code and will return a JSON object of graphemes and phonemes recognized and their locations. The language code will need to be verified against a list of languages with available grapheme to phoneme models and those without a matching model will be rejected. The transcript will need to be saved to a file before running the Montreal Forced Aligner. Results are output in a proprietary format, I've got scripts on my local machine that can format it to JSON that I'll provide you. A large part of this task will be getting the MFA and its associated models accessible.
We need to have an API endpoint up for grapheme and phoneme detection. This isn't provided by any cloud service that I am aware of and we're currently handling it through a tool called the Montreal Forced Aligner. I'm thinking a Google Cloud function that calls a Flask server that we manage.