Vietnamese ASR sequence-to-sequence model. This model supports output normalizing text, labeling timestamps, and segmenting multiple speakers.

# !pip install transformers, sentencepiece

from transformers import SpeechEncoderDecoderModel
from transformers import AutoFeatureExtractor, AutoTokenizer, GenerationConfig
import torchaudio
import torch

model_path = 'nguyenvulebinh/wav2vec2-bartpho'
model = SpeechEncoderDecoderModel.from_pretrained(model_path).eval()
feature_extractor = AutoFeatureExtractor.from_pretrained(model_path)
tokenizer = AutoTokenizer.from_pretrained(model_path)
if torch.cuda.is_available():
  model = model.cuda()


def decode_tokens(token_ids, skip_special_tokens=True, time_precision=0.02):
    timestamp_begin = tokenizer.vocab_size
    outputs = [[]]
    for token in token_ids:
        if token >= timestamp_begin:
            timestamp = f" |{(token - timestamp_begin) * time_precision:.2f}| "
            outputs.append(timestamp)
            outputs.append([])
        else:
            outputs[-1].append(token)
    outputs = [
        s if isinstance(s, str) else tokenizer.decode(s, skip_special_tokens=skip_special_tokens) for s in outputs
    ]
    return "".join(outputs).replace("< |", "<|").replace("| >", "|>")

def decode_wav(audio_wavs, asr_model, prefix=""):
  device = next(asr_model.parameters()).device
  input_values = feature_extractor.pad(
    [{"input_values": feature} for feature in audio_wavs],
    padding=True,
    max_length=None,
    pad_to_multiple_of=None,
    return_tensors="pt",
  )

  output_beam_ids = asr_model.generate(
    input_values['input_values'].to(device), 
    attention_mask=input_values['attention_mask'].to(device),
    decoder_input_ids=tokenizer.batch_encode_plus([prefix] * len(audio_wavs), return_tensors="pt")['input_ids'][..., :-1].to(device),
    generation_config=GenerationConfig(decoder_start_token_id=tokenizer.bos_token_id),
    max_length=250, 
    num_beams=25, 
    no_repeat_ngram_size=4, 
    num_return_sequences=1, 
    early_stopping=True,
    return_dict_in_generate=True,
    output_scores=True,
  )

  output_text = [decode_tokens(sequence) for sequence in output_beam_ids.sequences]

  return output_text


# https://huggingface.co/nguyenvulebinh/wav2vec2-bartpho/resolve/main/sample_news.wav
print(decode_wav([torchaudio.load('sample_news.wav')[0].squeeze()], model))

# <|0.00| Gia đình cho biết, nhiều lần đã từng gọi điện báo chính quyền và lực lượng an ninh địa phương nhưng đều không có tác dụng |7.00|>
# <|8.14| Không ai giúp đỡ được mình một chút nào cả, nên là lúc đó là lúc tuyệt vọng nhất, nó tra tấn mình cực kỳ khổ, gây cái tâm lý ức chế rất là nhiều, rất là lớn |19.02|>

Citation

This repository uses the idea from the following paper. Please cite the paper if this model is used to help produce published results or is incorporated into other software.

@INPROCEEDINGS{10446589,
  author={Nguyen, Thai-Binh and Waibel, Alexander},
  booktitle={ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, 
  title={Synthetic Conversations Improve Multi-Talker ASR}, 
  year={2024},
  volume={},
  number={},
  pages={10461-10465},
  keywords={Systematics;Error analysis;Knowledge based systems;Oral communication;Signal processing;Data models;Acoustics;multi-talker;asr;synthetic conversation},
  doi={10.1109/ICASSP48485.2024.10446589}
}

Contact

[email protected]

Follow

Downloads last month
4,274
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.