Automatic Speech Recognition
Transformers
PyTorch
Swiss German
wav2vec2
hf-asr-leaderboard
robust-speech-event
Eval Results (legacy)
# Load model directly
from transformers import AutoProcessor, AutoModelForCTC
processor = AutoProcessor.from_pretrained("manifoldix/xlsr-sg-lm")
model = AutoModelForCTC.from_pretrained("manifoldix/xlsr-sg-lm")Quick Links
XLSR-1b Swiss German
Fine-tuned on the Swiss parliament dataset from FHNW v1 (70h).
Tested on the Swiss parliament test set with a WER of 34.6%
Tested on the "Swiss German Dialects" with a WER of 40%
Both test sets can be accessed here: fhnw_datasets
The Swiss German dialect private test set has been uploaded on huggingface: huggingface_swiss_dialects
- Downloads last month
- 7
Evaluation results
- Test WER on Swiss parliamentself-reported34.6%
- Test WER on Swiss dialect test setself-reported40%
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="manifoldix/xlsr-sg-lm")