wav2vec2-large-chinese-zh-cn

Fine-tuned XLSR-53 large model for speech recognition in Chinese

Fine-tuned facebook/wav2vec2-large-xlsr-53 on Chinese using the train and validation splits of Common Voice 6.1, CSS10 and ST-CMDS. When using this model, make sure that your speech input is sampled at 16kHz.

This model has been fine-tuned on RTX3090 for 50h

The script used for training can be found here: https://github.com/jonatasgrosman/wav2vec2-sprint

Usage

The model can be used directly (without a language model) as follows...

Using the HuggingSound library:

from huggingsound import SpeechRecognitionModel
model = SpeechRecognitionModel("wbbbbb/wav2vec2-large-chinese-zh-cn")
audio_paths = ["/path/to/file.mp3", "/path/to/another_file.wav"]
transcriptions = model.transcribe(audio_paths)

Evaluation

The model can be evaluated as follows on the Chinese (zh-CN) test data of Common Voice.

import torch
import re
import librosa
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import warnings
import os

os.environ["KMP_AFFINITY"] = ""


LANG_ID = "zh-CN"
MODEL_ID = "zh-CN-output-aishell"
DEVICE = "cuda"

test_dataset = load_dataset("common_voice", LANG_ID, split="test")

wer = load_metric("wer")
cer = load_metric("cer")



processor = Wav2Vec2Processor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
model.to(DEVICE)

# Preprocessing the datasets.
# We need to read the audio files as arrays
def speech_file_to_array_fn(batch):
    with warnings.catch_warnings():
        warnings.simplefilter("ignore")
        speech_array, sampling_rate = librosa.load(batch["path"], sr=16_000)
    batch["speech"] = speech_array
    batch["sentence"] = (
        re.sub("([^\u4e00-\u9fa5\u0030-\u0039])", "", batch["sentence"]).lower() + " "
    )
    return batch


test_dataset = test_dataset.map(
    speech_file_to_array_fn,
    num_proc=15,
    remove_columns=['client_id', 'up_votes', 'down_votes', 'age', 'gender', 'accent', 'locale', 'segment'],
)

# Preprocessing the datasets.
# We need to read the audio files as arrays
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(DEVICE),
            attention_mask=inputs.attention_mask.to(DEVICE),
        ).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)

predictions = [x.lower() for x in result["pred_strings"]]
references = [x.lower() for x in result["sentence"]]

print(
    f"WER: {wer.compute(predictions=predictions, references=references, chunk_size=1000) * 100}"
)
print(f"CER: {cer.compute(predictions=predictions, references=references) * 100}")

Test Result:

In the table below I report the Word Error Rate (WER) and the Character Error Rate (CER) of the model. I ran the evaluation script described above on other models as well (on 2022-07-18). Note that the table below may show different results from those already reported, this may have been caused due to some specificity of the other evaluation scripts used.

Citation

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

@misc{grosman2021xlsr53-large-chinese,
  title={Fine-tuned {XLSR}-53 large model for speech recognition in {C}hinese},
  author={Grosman, Jonatas},
  howpublished={\url{https://huggingface.co/wbbbbb/wav2vec2-large-chinese-zh-cn}},
  year={2021}
}