@inproceedings{barmandah-etal-2026-ketaba,
    title = "Ketaba-{OCR} at {AR}-{MS} {N}akba{NLP} 2026: Efficient Adaptation of Vision-Language Models for Handwritten Recognition",
    author = "Barmandah, Hassan  and
      Eldin, Fatimah Emad  and
      Al Jallad, Khloud  and
      Nacar, Omer",
    editor = "Jarrar, Mustafa  and
      El-Haj, Mo  and
      Haddad, Amal  and
      Atiani, Serin  and
      Abudalfa, Shadi  and
      Regier, Terry  and
      Rayson, Paul  and
      Sima{'}an, Khalil  and
      Mansour, Camille",
    booktitle = "Proceedings of the 2nd International Workshop on Nakba Narratives as Language Resources @ {LREC} 2026",
    month = may,
    year = "2026",
    address = "Palma, Mallorca (Spain)",
    publisher = "ELRA Language Resources Association (ELRA)",
    url = "https://preview.aclanthology.org/revision-previews/2026.nakbanlp-1.21/",
    doi = "10.63317/3inc2znes52o",
    pages = "160--170",
    abstract = "This paper presents Ketaba-OCR-LoRA, a system developed for the NakbaNLP 2026 Shared Task on Arabic Manuscript Understanding (Subtask 2), which targets the transcription of the historically significant Omar Al-Saleh Memoir Collection written in Ruq{'}ah and Naskh scripts. We propose a parameter-efficient adaptation of a publicly available pretrained Arabic-English Handwritten Text Recognition (HRT) model, originally trained on handwritten corpora including the Muharaf dataset. Instead of adapting general Vision-Language Models from scratch, we fine-tune the HRT backbone using Low-Rank Adaptation (LoRA) and 4-bit quantization (QLoRA), reducing memory requirements from 40GB to approximately 8GB. Our final submission combines multiple model variants through a novel Linear+Boost weighted ensemble strategy. Our approach achieves a CER of 0.0819 and WER of 0.2588 on the blind test set (per-line evaluation), ranking 1st on per-line evaluation; on the official corpus-wide leaderboard, we rank 3rd (CER 0.0938, WER 0.2996). This work demonstrates that specialized pretrained HRT models substantially outperform general-purpose Vision-Language Models for Arabic manuscript transcription, and that parameter-efficient fine-tuning provides a practical and reproducible approach for low-resource cultural heritage digitization."
}