Hello everyone, submitted by \/u\/youcefotmani <\/a>
We’re developing an AI-powered app, Faseeh AI, that detects pronunciation mistakes in Quranic recitation and gives users precise, real-time feedback.
Our current model was trained on hundreds of hours of professional recitations and high-quality, clean audio from well-known reciters. The model performs well on similar input, but struggles with real-world users: different accents, non-native speakers, beginners, children, women, and anyone who doesn’t sound like a professional reciter.
To fix this, we need to train on diverse, real-world recitation data not studio-quality professional audio, but recordings that reflect how actual learners sound.
Specifically, we’re looking for:<\/strong>
– Recitation datasets from non-professional or everyday users
– Diverse demographics: male\/female, kids\/adults, beginner\/intermediate
– Multiple accents and mother tongues (Malay, Indonesian, Urdu, English, Turkish, etc.)
– Any publicly available or research-use datasets we may have missed
We’ve already explored academic sources, but still not enough.
If you know of any dataset, research project, university study, or community effort collecting this type of audio, we would genuinely appreciate the lead.
We’re also open to ethical data collection partnerships if any researchers or institutions are working in this space.
Happy to share more about the project if helpful.
Thank you very much in advance.<\/p>\n<\/div>\n
[link]<\/a><\/span> [comments]<\/a><\/span><\/p>
