How we can train languages other than Korean?

[usamaehsan]

No description provided.

[sooftware]

If you have dataset (audio & transcript), You can train, but you have to modify a few things.

1. First, you should make a character label like this.
2. Second, you have to convert transcripts to ids on character labels.
3. The code should be modified considering the structure of the dataset. (You will need to modify the kospech/data directory.)

If you have any other questions, please leave an issue at any time.

[usamaehsan]

@sooftware can you please tell whats freq in aihub_labels.csv?

[sooftware]

Freq is not very important. Freq (frequency) is a count of how often it appeared in a dataset.
Freq is not used in training, so you don't have to worry.

[usamaehsan]

thanks,you are very helpful
i am new to this feild and there is no speech recognition available for my language urdu thats why i am trying to build urdu speech recognition system using this repo

[sooftware]

It's a great honor. I want to help you as much as I can, so please leave an issue or contact me at sh951011@gmail.com at any time.
During the week, I may be busy and answer late, but I will answer as soon as possible on the weekend.

[usamaehsan]

thanks