How to use both pronunciation file and a structured file in custom speech to text and speech studio?

TR Ganesh 0

I am using Microsoft's speech studio's Custom speech to text service. I have created a project and when uploading the data files there are different format in which I can upload. For my project I am using multiple formats of data, I want to use both pronunciation file and a structure defined for my project but when I am selecting them it is showing error that "cannot use both pronunciation file and structured file". Can anyone please tell me how to use both for training the Speech Studio Custom Speech Model.

dupammi 8,465 Reputation points Microsoft Vendor

2024-07-12T07:58:49.71+00:00

Hi @TR Ganesh

Thank you for your question.

Based on the documentation for Microsoft's Speech Studio, it's currently not possible to use both a pronunciation file and a structured text phonetic pronunciation file simultaneously for training a custom speech model. The system does not support combining these datasets.

You can either use a pronunciation data file on its own, or you can add pronunciation within a structured text data file. The Speech service doesn't support training a model with both of those datasets as input.

You have two options:

Use Only Pronunciation File: You can upload a pronunciation file separately. This file will train the model on how specific words sound.

Use Structured Text Phonetics: Alternatively, you can use a structured text file that specifies phonetic pronunciations per syllable using the Universal Phone Set (UPS).

To proceed with your project, you'll need to choose one of these approaches based on your specific needs. Unfortunately, combining both types of data files is not supported at this time.I hope you understand. Thank you.

Share via

How to use both pronunciation file and a structured file in custom speech to text and speech studio?

Your answer