Is it possible to train a custom vocabulary for non english transcription?

[jzju]

Can you customize the language model with your own words? Something like creating an https://github.com/kpu/kenlm model for wav2vec2.

[jongwook]

Please see #328 (comment)