Non pretrained languages

Hello,
I am wondering for Document understanding. For languages other than pre-trained ones
Is it hard to train? I have being looking through the documentation and this forum and am not sure where to start.

@Snorri_Arinbjarnar

You only need to label them and show what the data looks like so that ml can identify the characters you need…

Datalabeling for a good number of dataset is recommended (atleast 10 documents per type)

Cheers