Thaka at KSAA-2026 Task 2: Regularized Fine-Tuning for Arabic Speech Diacritization
Fuente:
arXiv
Salvato in:
| Autori principali: | , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910255431024640 |
|---|---|
| author | Alamr, Meshal Alqaeri, Hassan Aldahlawi, Abdullah |
| author_facet | Alamr, Meshal Alqaeri, Hassan Aldahlawi, Abdullah |
| contents | We describe the winning system for Task 2 of the KSAA-2026 Shared Task on Arabic Speech Dictation with Automatic Diacritization. The task requires producing fully diacritized Arabic text from speech audio and undiacritized transcripts, with only 2,327 training samples available and no external data permitted. Our system fine-tunes CATT-Whisper, a character-level multimodal model combining a pretrained CATT text encoder with a frozen Whisper speech encoder. The key to our approach is training regularization: R-Drop consistency regularization, Optuna-optimized hyperparameters with high weight decay, and Focal Loss. At inference, we average 200 stochastic forward passes across four model checkpoints using Monte Carlo Dropout at the softmax probability level. The system achieves 23.26% WER on the primary leaderboard metric (with case endings, including no-diacritic positions), placing 1st among all participants. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_25928 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Thaka at KSAA-2026 Task 2: Regularized Fine-Tuning for Arabic Speech Diacritization Alamr, Meshal Alqaeri, Hassan Aldahlawi, Abdullah Computation and Language Sound Audio and Speech Processing I.2.7; I.5.4 We describe the winning system for Task 2 of the KSAA-2026 Shared Task on Arabic Speech Dictation with Automatic Diacritization. The task requires producing fully diacritized Arabic text from speech audio and undiacritized transcripts, with only 2,327 training samples available and no external data permitted. Our system fine-tunes CATT-Whisper, a character-level multimodal model combining a pretrained CATT text encoder with a frozen Whisper speech encoder. The key to our approach is training regularization: R-Drop consistency regularization, Optuna-optimized hyperparameters with high weight decay, and Focal Loss. At inference, we average 200 stochastic forward passes across four model checkpoints using Monte Carlo Dropout at the softmax probability level. The system achieves 23.26% WER on the primary leaderboard metric (with case endings, including no-diacritic positions), placing 1st among all participants. |
| title | Thaka at KSAA-2026 Task 2: Regularized Fine-Tuning for Arabic Speech Diacritization |
| topic | Computation and Language Sound Audio and Speech Processing I.2.7; I.5.4 |
| url | https://arxiv.org/abs/2605.25928 |