Quranic recitation

[maq796113]

I thinking about using it to make a real-time live feedback on Quranic recitation using Graphene to Phoneme converters...

[OmarWKH]

Have a look at https://www.tarteel.ai/ for inspiration

[maq796113]

I have looked into it

[Ashraf-Ali-aa]

@maq796113 have a look at https://github.com/tarekeldeeb/DeepSpeech-Quran it uses tarteel.ai data models, I was looking into using the data from tarteel to train whisper to transcribe Quran audio, using https://huggingface.co/blog/fine-tune-whisper

[maq796113]

I am using Brain Speech

[Ashraf-Ali-aa]

@maq796113 if you're interested we can work together, I am trying to train whisper on quran audio

[maq796113]

Is there way I can contact you?
For now I didn't do much, I am working for senior project. I didn't do much except for make use of Natural Language Toolkit to find the BiLingual Evaluation Understudy score....
https://colab.research.google.com/drive/1j2eta-5ySTks0tZvKnMvmZCCEDrAKuPo?authuser=2#scrollTo=eG5vTiJgYsRi

you can go over this colab

[Ashraf-Ali-aa]

@maq796113 message me by joining the quran.com discord server https://discord.gg/FxRWSBfWxn or using my username Ashraf.Ali#0533 , also I can't access the colab URL

[maq796113]

I have contacted you on discord

[abodacs]

@maq796113
I can't access the colab URL
can you fix it?

Thanks

[BasselAshi]

I'm very late to the party but I am looking for a model to run for transcribing Quran and I am not sure where to get started. LLM models are a gray area for me.

[msis]

Tarteel put out datasets and a couple of models out.

Check them the org page on HF: https://huggingface.co/tarteel-ai

[brilliant-ember]

I found this dataset for Quran recitations by different readers, this version is resampled to 16khz https://www.openslr.org/132/
And this version is the original audio at various sampling rates: https://www.kaggle.com/datasets/bigguyubuntu/quran-ayat-speech-to-text
It's a very good dataset, I had problems with their "tsv" files so I created my own Arabic labels and put it in a csv, it was easy because they have the ayat numbered in the dataset.

[maq796113]

how are you planning on using this dataset?