README.md

metadata

language:
  - it
license: apache-2.0
datasets:
  - mozilla-foundation/common_voice_11_0
metrics:
  - wer
  - cer
tags:
  - audio
  - automatic-speech-recognition
  - hf-asr-leaderboard
  - it
  - mozilla-foundation/common_voice_11_0
  - speech
  - wav2vec2
model-index:
  - name: XLS-R Wav2Vec2 CV11Ita by radiogroup crits
    results:
      - task:
          name: Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Common Voice 11.0 italian
          type: mozilla-foundation/common_voice_11_0
          args: it
        metrics:
          - name: Test WER
            type: wer
            value: 7.12
          - name: Test CER
            type: cer
            value: 1.75
          - name: Test WER (+LM)
            type: wer
            value: 5.77
          - name: Test CER (+LM)
            type: cer
            value: 1.51

XLS-R-1B-CV11ITA-LMWIKI500

Fine-tuned XLS-R 1B model for speech recognition in Italian

Fine-tuned facebook/wav2vec2-xls-r-1b on Italian using the train and validation splits of Common Voice 11.0.

When using this model, make sure that your speech input is sampled at 16kHz.

Language model information

Our language model was generated using a 500-characters data set for each Italian Wikipedia article.

Download CommonVoice11.0 dataset for italian language

from datasets import load_dataset

dataset = load_dataset("mozilla-foundation/common_voice_11_0", "it", use_auth_token=True)

Evaluation Commands

To evaluate on mozilla-foundation/common_voice_11_0 with split test:

python eval.py --model_id radiogroup-crits/wav2vec2-xls-r-1b-cv11ita-lmwiki500 --dataset mozilla-foundation/common_voice_11_0 --config it --split test --log_outputs --greedy

mv log_mozilla-foundation_common_voice_11_0_it_test_predictions.txt log_mozilla-foundation_common_voice_11_0_it_test_predictions_greedy.txt

mv log_mozilla-foundation_common_voice_11_0_it_test_targets.txt log_mozilla-foundation_common_voice_11_0_it_test_targets_greedy.txt

mv mozilla-foundation_common_voice_11_0_it_test_eval_results.txt mozilla-foundation_common_voice_11_0_it_test_eval_results_greedy.txt

python eval.py --model_id radiogroup-crits/wav2vec2-xls-r-1b-cv11ita-lmwiki500 --dataset mozilla-foundation/common_voice_11_0 --config it --split test --log_outputs

mv log_mozilla-foundation_common_voice_11_0_it_test_predictions.txt log_mozilla-foundation_common_voice_11_0_it_test_predictions_lm.txt

mv log_mozilla-foundation_common_voice_11_0_it_test_targets.txt log_mozilla-foundation_common_voice_11_0_it_test_targets_lm.txt

mv mozilla-foundation_common_voice_11_0_it_test_eval_results.txt mozilla-foundation_common_voice_11_0_it_test_eval_results_lm.txt

Citation

If you want to cite this model you can use this:

@misc{crits2023wav2vec2-xls-r-1b-cv11ita-lmwiki500,
  title={XLS-R Wav2Vec2 CV11Ita by radiogroup crits},
  author={Teraoni Prioletti Raffaele, Casagranda Paolo and Russo Francesco},
  publisher={Hugging Face},
  journal={Hugging Face Hub},
  howpublished={\url{https://huggingface.co./radiogroup-crits/wav2vec2-xls-r-1b-cv11ita-lmwiki500}},
  year={2023}
}