语言: 日语
数据集:

• common_voice
  评估指标:
• wer
  标签:
• 音频
• 自动语音识别
• 语音
• xlsr微调周
  许可证: apache-2.0
  模型索引:
• 名称: XLSR Wav2Vec2 日语平假名 by Chien Vu
  结果:
  • 任务:
    名称: 语音识别
    类型: automatic-speech-recognition
    数据集:
    名称: Common Voice 日语
    类型: common_voice
    参数: ja
    指标:
    • 名称: 测试WER
      类型: wer
      值: 24.74
    • 名称: 测试CER
      类型: cer
      值: 10.99

Wav2Vec2-Large-XLSR-53-日语

基于facebook/wav2vec2-large-xlsr-53模型，使用Common Voice和东京大学Saruwatari实验室的日语语音语料库JSUT进行微调。
使用该模型时，请确保语音输入采样率为16kHz。

使用方法

该模型可直接使用（无需语言模型），如下所示：

!pip install mecab-python3  
!pip install unidic-lite  
!pip install pykakasi  
!python -m unidic download  
import torch  
import torchaudio  
import librosa  
from datasets import load_dataset  
import MeCab  
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor  
import re  
# 配置  
wakati = MeCab.Tagger("-Owakati")  
chars_to_ignore_regex = '[\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\,\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\、\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\。\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\．\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\「\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\」\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\…\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\？\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\・]'  
kakasi = pykakasi.kakasi()  
kakasi.setMode("J","H")  
kakasi.setMode("K","H")  
kakasi.setMode("r","Hepburn")  
conv = kakasi.getConverter()  
# 加载数据、处理器和模型  
test_dataset = load_dataset("common_voice", "ja", split="test[:2%]")  
processor = Wav2Vec2Processor.from_pretrained("vumichien/wav2vec2-large-xlsr-japanese-hỉragana")  
model = Wav2Vec2ForCTC.from_pretrained("vumichien/wav2vec2-large-xlsr-japanese-hỉragana")  
resampler = lambda sr, y: librosa.resample(y.numpy().squeeze(), sr, 16_000)  
# 数据集预处理  
def speech_file_to_array_fn(batch):  
    batch["sentence"] = conv.do(wakati.parse(batch["sentence"]).strip())  
    batch["sentence"] = re.sub(chars_to_ignore_regex,'', batch["sentence"]).strip()  
    speech_array, sampling_rate = torchaudio.load(batch["path"])  
    batch["speech"] = resampler(sampling_rate, speech_array).squeeze()  
    return batch  
test_dataset = test_dataset.map(speech_file_to_array_fn)  
inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)  
with torch.no_grad():  
    logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits  
predicted_ids = torch.argmax(logits, dim=-1)  
print("预测:", processor.batch_decode(predicted_ids))  
print("参考:", test_dataset["sentence"][:2])  

评估

该模型可在Common Voice的日语测试数据上进行如下评估：

!pip install mecab-python3  
!pip install unidic-lite  
!pip install pykakasi  
!python -m unidic download  
import torch  
import librosa  
import torchaudio  
from datasets import load_dataset, load_metric  
import MeCab  
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor  
import re  
# 配置  
wakati = MeCab.Tagger("-Owakati")  
chars_to_ignore_regex = '[\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\,\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\、\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\。\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\．\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\「\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\」\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\…\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\？\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\・]'  
kakasi = pykakasi.kakasi()  
kakasi.setMode("J","H")  
kakasi.setMode("K","H")  
kakasi.setMode("r","Hepburn")  
conv = kakasi.getConverter()  
# 加载数据、处理器和模型  
test_dataset = load_dataset("common_voice", "ja", split="test")  
wer = load_metric("wer")  
cer = load_metric("cer")  
processor = Wav2Vec2Processor.from_pretrained("vumichien/wav2vec2-large-xlsr-japanese-hỉragana")  
model = Wav2Vec2ForCTC.from_pretrained("vumichien/wav2vec2-large-xlsr-japanese-hỉragana")  
model.to("cuda")  
resampler = lambda sr, y: librosa.resample(y.numpy().squeeze(), sr, 16_000)  
# 数据集预处理  
def speech_file_to_array_fn(batch):  
    batch["sentence"] = conv.do(wakati.parse(batch["sentence"]).strip()  
    batch["sentence"] = re.sub(chars_to_ignore_regex,'', batch["sentence"]).strip()  
    speech_array, sampling_rate = torchaudio.load(batch["path"])  
    batch["speech"] = resampler(sampling_rate, speech_array).squeeze()  
    return batch  
test_dataset = test_dataset.map(speech_file_to_array_fn)  
# 评估函数  
def evaluate(batch):  
    inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)  
    with torch.no_grad():  
        logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits  
    pred_ids = torch.argmax(logits, dim=-1)  
    batch["pred_strings"] = processor.batch_decode(pred_ids)  
    return batch  
result = test_dataset.map(evaluate, batched=True, batch_size=8)  
print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))  
print("CER: {:2f}".format(100 * cer.compute(predictions=result["pred_strings"], references=result["sentence"])))  

测试结果

WER: 24.74%,
CER: 10.99%

训练

使用了Common Voice的train、validation数据集和日语语音语料库数据集进行训练。