@inproceedings{nadhir-daoud-anouar-ben-messaoud-2025-phoneme,
    title = "Phoneme-level mispronunciation detection in {Q}uranic recitation using {S}hallow{T}ransformer",
    author = "Daoud, Mohamed Nadhir  and
      Ben Messaoud, Mohamed Anouar",
    editor = "Darwish, Kareem  and
      Ali, Ahmed  and
      Abu Farha, Ibrahim  and
      Touileb, Samia  and
      Zitouni, Imed  and
      Abdelali, Ahmed  and
      Al-Ghamdi, Sharefah  and
      Alkhereyf, Sakhar  and
      Zaghouani, Wajdi  and
      Khalifa, Salam  and
      AlKhamissi, Badr  and
      Almatham, Rawan  and
      Hamed, Injy  and
      Alyafeai, Zaid  and
      Alowisheq, Areeb  and
      Inoue, Go  and
      Mrini, Khalil  and
      Alshammari, Waad",
    booktitle = "Proceedings of The Third Arabic Natural Language Processing Conference: Shared Tasks",
    month = nov,
    year = "2025",
    address = "Suzhou, China",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/author-page-yu-wang-polytechnic/2025.arabicnlp-sharedtasks.63/",
    doi = "10.18653/v1/2025.arabicnlp-sharedtasks.63",
    pages = "457--463",
    ISBN = "979-8-89176-356-2",
    abstract = "Preserving the integrity of Qur{'}anic recitation requires accurate pronunciation, as even subtle mispronunciations can alter meaning. Automatic assessment of Qur{'}anic recitation at the phoneme level is therefore a critical and challenging task. We present ShallowTransformer, a lightweight and computationally efficient transformer model leveraging Wav2vec2.0 features and trained with CTC loss for phoneme-level mispronunciation detection. Evaluated on the Iqra{'}Eval benchmark (QuranMB.v2), our model outperforms published BiLSTM baselines on QuranMB.v1 while achieving competitive performance relative to the official Iqra{'}Eval challenge baselines, which are not yet fully documented. Such improvements are particularly important in assisted Qur{'}an learning, as accurate phonetic feedback supports correct recitation and preserves textual integrity. These results highlight the effectiveness of transformer architectures in capturing subtle pronunciation errors while remaining deployable for practical applications."
}