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%