CSTR-Edinburgh/merlin

What is a question file and how to create it?

priyanthini opened this issue · 7 comments

I'm building a Tamil (indic language) voice using Merlin. I got a very bad output. I think it might be due to the question file. I just used the default question file. I want to know the purpose of the question file and how to create a question file for a new language.

Hello,

the question files are there: https://github.com/CSTR-Edinburgh/merlin/tree/master/misc/questions

You have examples of multiple languages.

For Merlin, the purpose of the question file is to create the input feature vector from the label file.

@seblemaguer Thank you so much for the reply. But I want to know how to create a question file for our own language?

Do you have a phoneme set defined?

@seblemaguer Yes I have. Is there any reference to create a question file?

Thank you.

Not really. The question file is created in relation to the label files you are extracting from the text. Merlin is using the convention proposed in HTS. So if you don't know how it is done, I would suggest having a look at this: https://wiki.inf.ed.ac.uk/twiki/pub/CSTR/F0parametrisation/hts_lab_format.pdf

However, to have a good starting point, you simply copy the English question file to a new question file and you replace the questions related to the phonetic information from the English to your phonemic set. They are the first questions you will see in the file.

@seblemaguer Okay I will refer those and check. Thank you so much.

@seblemaguer I'm unable to create a question file for Tamil language. What can I do to solve this problem?