The "faster-whisper-large-v3-ca-3catparla" is an acoustic model based on a faster-whisper version of projecte-aina/whisper-large-v3-ca-3catparla suitable for Automatic Speech Recognition in Catalan.
The "faster-whisper-large-v3-ca-3catparla" is the result of converting the projecte-aina/whisper-large-v3-ca-3catparla into a lighter model using a python module called faster-whisper.
The specific dataset used to create the projecte-aina/whisper-large-v3-ca-3catparla model is called "3CatParla" (soon to be published).
This model can used for Automatic Speech Recognition (ASR) in Catalan. The model is intended to transcribe audio files in Catalan to plain text without punctuation.
To see an updated and functional version of this code, please see our our Notebook.
In order to use this model, you may install faster-whisper
Create a virtual environment:
python -m venv /path/to/venv
Activate the environment:
source /path/to/venv/bin/activate
Install the modules:
pip install faster-whisper
In order to transcribe audio in Catalan using this model, you can follow this example:
from faster_whisper import WhisperModel
model_size = "projecte-aina/faster-whisper-large-v3-ca-3catparla"
# Run on GPU with FP16
model = WhisperModel(model_size, device="cuda", compute_type="float16")
# or run on GPU with INT8
#model = WhisperModel(model_size, device="cuda", compute_type="int8_float16")
# or run on CPU with INT8
# model = WhisperModel(model_size, device="cpu", compute_type="int8")
segments, info = model.transcribe("audio_in_catalan.mp3", beam_size=5, task="transcribe",language="ca")
print("Detected language '%s' with probability %f" % (info.language, info.language_probability))
for segment in segments:
print("[%.2fs -> %.2fs] %s" % (segment.start, segment.end, segment.text))
This model is not a direct result of training. It is a conversion of a Whisper model using faster-whisper. The procedure to create the model is as follows:
ct2-transformers-converter --model projecte-aina/whisper-large-v3-ca-3catparla
--output_dir faster-whisper-large-v3-ca-3catparla
--copy_files preprocessor_config.json
--quantization float16
If this model contributes to your research, please cite the work:
@inproceedings{hernandez20243catparla,
title={3CatParla: A New Open-Source Corpus of Broadcast TV in Catalan for Automatic Speech Recognition},
author={Hern{\'a}ndez Mena, Carlos Daniel and Armentano Oller, Carme and Solito, Sarah and K{\"u}lebi, Baybars},
booktitle={Proc. IberSPEECH 2024},
pages={176--180},
year={2024}
}
The conversion process was perform during July (2024) in the Language Technologies Unit of the Barcelona Supercomputing Center by Carlos Daniel Hernández Mena.
For further information, please send an email to langtech@bsc.es.
Copyright(c) 2024 by Language Technologies Unit, Barcelona Supercomputing Center.
This work has been promoted and financed by the Generalitat de Catalunya through the Aina project.
The conversion of the model was possible thanks to the compute time provided by Barcelona Supercomputing Center through MareNostrum 5.