Fact · Languages and accents
14.0percent word error rate
The Swiss German corpus authors fine-tuned a Wav2Vec2 XLS-R model on their data and reached a word error rate of 14.0 percent on their test set, as Dolev and colleagues report.
Checked against its source on October 17, 2026
In the source's own words
“They fine-tuned a Wav2Vec2 XLS-R model on the corpus and reached a WER of 14.0 and a BLEU score of 74.7 on their test set.”
Does Whisper Understand Swiss German? An Automatic, Qualitative and Human Evaluation, Eyal Liron Dolev et al., Proceedings of the 11th Workshop on NLP for Similar Languages, Varieties and Dialects (VarDial 2024), 2024, Section 2, related work on Swiss German corpora.
Open the original- What it measures
- This is a model trained on Swiss German data, which a general model such as Whisper is not. Dolev and colleagues cite the result and do not rerun it.
- What it does not show
- Reported by a second paper. The figure is for the corpus test set, and it is not a measure of everyday dictation.