README.md

metadata

license: apache-2.0
tags:
  - generated_from_trainer
metrics:
  - wer
  - cer
model-index:
  - name: uniTKU-hubert-japanese-asr
    results:
      - task:
          name: Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: common_voice_11_0
          type: common_voice
          args: ja
        metrics:
          - name: Test WER
            type: wer
            value: 27.447168
          - name: Test CER
            type: cer
            value: 11.607944
datasets:
  - mozilla-foundation/common_voice_11_0
language:
  - ja

uniTKU-hubert-japanese-asr

This model was fine-tuned on a dataset provided by uniTKU, and it has maintained the original performance metrics on the common_voice_11_0 dataset.

This model can only predict Hiragana.

Training Procedure

Fine-tuning on the uniTKU dataset led to the following results:

Step	Training Loss	Validation Loss	WER
100	0.910100	1.051628	0.669118
200	0.747700	0.691642	0.551471
300	0.718000	0.705763	0.544118
400	0.663700	0.532831	0.397059
500	0.667700	0.491024	0.352941
600	0.546800	0.365637	0.330882
700	0.569000	0.274410	0.279412
800	0.591800	0.274801	0.235294
900	0.575400	0.257891	0.220588
1000	0.579100	0.280559	0.205882

Training hyperparameters

The training hyperparameters remained consistent throughout the fine-tuning process:

• learning_rate: 1e-4
• train_batch_size: 16
• eval_batch_size: 16
• gradient_accumulation_steps: 2
• num_train_epochs: 15
• lr_scheduler_type: linear

How to evaluate the model

from transformers import HubertForCTC, Wav2Vec2Processor
from datasets import load_dataset
import torch
import torchaudio
import librosa
import numpy as np
import re
import MeCab
import pykakasi
from evaluate import load

model = HubertForCTC.from_pretrained('TKU410410103/uniTKU-hubert-japanese-asr')
processor = Wav2Vec2Processor.from_pretrained("TKU410410103/uniTKU-hubert-japanese-asr")

# load dataset
test_dataset = load_dataset('mozilla-foundation/common_voice_11_0', 'ja', split='test')
remove_columns = [col for col in test_dataset.column_names if col not in ['audio', 'sentence']]
test_dataset = test_dataset.remove_columns(remove_columns)

# resample
def process_waveforms(batch):
    speech_arrays = []
    sampling_rates = []

    for audio_path in batch['audio']:
        speech_array, _ = torchaudio.load(audio_path['path'])
        speech_array_resampled = librosa.resample(np.asarray(speech_array[0].numpy()), orig_sr=48000, target_sr=16000)
        speech_arrays.append(speech_array_resampled)
        sampling_rates.append(16000)

    batch["array"] = speech_arrays
    batch["sampling_rate"] = sampling_rates

    return batch

# hiragana
CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", "；", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
          "؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
          "{", "}", "=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。",
          "、", "﹂", "﹁", "‧", "～", "﹏", "，", "｛", "｝", "（", "）", "［", "］", "【", "】", "‥", "〽",
          "『", "』", "〝", "〟", "⟨", "⟩", "〜", "：", "！", "？", "♪", "؛", "/", "\\", "º", "−", "^", "'", "ʻ", "ˆ"]
chars_to_ignore_regex = f"[{re.escape(''.join(CHARS_TO_IGNORE))}]"

wakati = MeCab.Tagger("-Owakati")
kakasi = pykakasi.kakasi()
kakasi.setMode("J","H")
kakasi.setMode("K","H")
kakasi.setMode("r","Hepburn")
conv = kakasi.getConverter()

def prepare_char(batch):
    batch["sentence"] = conv.do(wakati.parse(batch["sentence"]).strip())
    batch["sentence"] = re.sub(chars_to_ignore_regex,'', batch["sentence"]).strip()
    return batch


resampled_eval_dataset = test_dataset.map(process_waveforms, batched=True, batch_size=50, num_proc=4)
eval_dataset = resampled_eval_dataset.map(prepare_char, num_proc=4)

# begin the evaluation process
wer = load("wer")
cer = load("cer")

def evaluate(batch):
    inputs = processor(batch["array"], 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

columns_to_remove = [column for column in eval_dataset.column_names if column != "sentence"]
batch_size = 16
result = eval_dataset.map(evaluate, remove_columns=columns_to_remove, batched=True, batch_size=batch_size)

wer_result = wer.compute(predictions=result["pred_strings"], references=result["sentence"])
cer_result = cer.compute(predictions=result["pred_strings"], references=result["sentence"])

print("WER: {:2f}%".format(100 * wer_result))
print("CER: {:2f}%".format(100 * cer_result))

Test results

The final model was evaluated as follows:

On uniTKU Dataset:

• WER: 20.5%

On common_voice_11_0:

• WER: 27.447168%
• CER: 11.607944%

Framework versions

• Transformers 4.39.1
• Pytorch 2.2.1+cu118
• Datasets 2.17.1