Tatar language data quality issues

[rsabirov]

Hello,

Where is this data for Tatar language is coming from?

I see a lot of garbage there, I barely found a Tatar words here.

I would like to improve this.

1. do you have some page with guidance how to train the model?
2. once I train it, should I create a PR with just model itself to that repo? where are storing raw data for training?

follow up for tesseract-ocr/langdata#305