ArTST-v3 (ASR task)

ArTST model finetuned for automatic speech recognition (speech-to-text) on MGB2 (best for Modern Standard Arabic)

Model Description

• Developed by: Speech Lab, MBZUAI
• Model type: SpeechT5
• Language: Arabic

How to Get Started with the Model

import soundfile as sf
from transformers import (
 SpeechT5Config,
 SpeechT5FeatureExtractor,
 SpeechT5ForSpeechToText,
 SpeechT5Processor,
 SpeechT5Tokenizer,
)


device = "cuda" if torch.cuda.is_available() else "CPU"

model_id="mbzuai/artst_asr_v3"

tokenizer = SpeechT5Tokenizer.from_pretrained(model_id)
processor = SpeechT5Processor.from_pretrained(model_id , tokenizer=tokenizer)
model = SpeechT5ForSpeechToText.from_pretrained(model_id).to(device)

audio, sr = sf.read("audio.wav")

inputs = processor(audio=audio, sampling_rate=sr, return_tensors="pt")
predicted_ids = model.generate(**inputs.to(device), max_length=150, num_beams=10)

transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)
print(transcription[0])

or using pipeline

import librosa
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor


model_id="mbzuai/artst_asr_v3"
processor = AutoProcessor.from_pretrained(model_id)
model = AutoModelForSpeechSeq2Seq.from_pretrained(model_id).to(device)
pipe = pipeline(
 "automatic-speech-recognition",
 model=model,
 tokenizer=processor.tokenizer,
 feature_extractor=processor.feature_extractor,
 torch_dtype=torch_dtype,
 device=device,
)

wav, sr = librosa.load("audio.wav", sr=16000)
pipe(wav, generate_kwargs={'num_beams': 10, 'early_stopping': True})['text']

Model Sources [optional]

• Repository: github
• Paper : ArXiv

Citation

BibTeX:

@misc{djanibekov2024dialectalcoveragegeneralizationarabic,
 title={Dialectal Coverage And Generalization in Arabic Speech Recognition}, 
 author={Amirbek Djanibekov and Hawau Olamide Toyin and Raghad Alshalan and Abdullah Alitr and Hanan Aldarmaki},
 year={2024},
 eprint={2411.05872},
 archivePrefix={arXiv},
 primaryClass={cs.CL},
 url={https://arxiv.org/abs/2411.05872}, 
}

@inproceedings{toyin-etal-2023-artst,
 title = "{A}r{TST}: {A}rabic Text and Speech Transformer",
 author = "Toyin, Hawau and
 Djanibekov, Amirbek and
 Kulkarni, Ajinkya and
 Aldarmaki, Hanan",
 booktitle = "Proceedings of ArabicNLP 2023",
 month = dec,
 year = "2023",
 address = "Singapore (Hybrid)",
 publisher = "Association for Computational Linguistics",
 url = "https://aclanthology.org/2023.arabicnlp-1.5",
 doi = "10.18653/v1/2023.arabicnlp-1.5",
 pages = "41--51",
}