Edit model card

🇺🇦 Join Ukrainian Speech Recognition Community - https://t.me/speech_recognition_uk

⭐ See other Ukrainian models - https://github.com/egorsmkv/speech-recognition-uk

This model has been trained on noisy data in order to make the acoustic model robust to noisy audio data.

This model has apostrophes and hyphens.

The language model is trained on the texts of the Common Voice dataset, which is used during training.

Special thanks for noised data to Dmytro Chaplynsky, https://lang.org.ua

Noisy dataset:

Metrics:

Dataset CER WER
CV10 (no LM) 0.0515 0.2617
CV10 (with LM) 0.0148 0.0524

Metrics on noisy data with standard model:

Dataset CER WER
CV10 (no LM) 0.1064 0.3926
CV10 (with LM) 0.0497 0.1265

More:

Downloads last month
10
Hosted inference API
or or
This model can be loaded on the Inference API on-demand.

Dataset used to train Yehor/wav2vec2-xls-r-300m-uk-with-small-lm-noisy