README.md

metadata

language:
  - ja
license: apache-2.0
tags:
  - audio
  - automatic-speech-recognition
  - speech
datasets:
  - Japanese accent datasets
metrics:
  - wer
model-index:
  - name: Wav2vec2 Accent Japanese
    results:
      - task:
          type: Speech Recognition
          name: automatic-speech-recognition
        dataset:
          type: accent_voice
          name: Japanese accent datasets
          args: ja
        metrics:
          - type: wer
            value: 15.82
            name: Test WER

Wav2Vec2 Accent Japanese

Fine-tuned facebook/wav2vec2-large-xlsr-53 on Japanese accent dataset When using this model, make sure that your speech input is sampled at 16kHz.

Test Result

WER: 15.82%