Not Gemma at AR-MS NakbaNLP 2026: Mubsir OCR: End-to-End Recognition of Arabic Handwritten Text

Ali Adel Ali, Mona Khaled Ali, Mohamed Emad Sayed, Ibrahim Naser Mostafa


Abstract
Historical Arabic handwritten OCR is difficult because of cursive script, fine diacritics, mixed numerals, and degraded media; classical segmentation pipelines compound errors, whereas end-to-end vision-language models can adapt when fine-tuned on in-domain data. We present Mubsir OCR, a systematic evaluation on the NAKBA dataset: an annotated set (15,962 training line crops and 2,095 val lines with ground truth, used for all nine experiments) and a separate blind AR-MS (Subtask 2) set (2,671 images; scores only via official submission). We compare external vs. in-house VLMs (Qwen2.5-VL 3B, Qwen3-VL-4B-Instruct, Gemma3), inference backends (vLLM/bf16 vs. HuggingFace/bf16), training length (16 vs. 32 epochs), and test-time preprocessing (CLAHE+unsharp). Best on the annotated val set: 8.59% CER / 25.87% WER (HuggingFace bf16); the same configuration attains 11.00% CER / 31.26% WER on the blind set. Domain-specific fine-tuning beats general-purpose checkpoints; preprocessing helps only marginally and is not recommended without train-time augmentation.
Anthology ID:
2026.nakbanlp-1.42
Volume:
Proceedings of the 2nd International Workshop on Nakba Narratives as Language Resources @ LREC 2026
Month:
May
Year:
2026
Address:
Palma, Mallorca (Spain)
Editors:
Mustafa Jarrar, Mo El-Haj, Amal Haddad, Serin Atiani, Shadi Abudalfa, Terry Regier, Paul Rayson, Khalil Sima’an, Camille Mansour
Venues:
NakbaNLP | WS
SIG:
Publisher:
ELRA Language Resources Association (ELRA)
Note:
Pages:
273–278
Language:
External URL:
https://lrec.elra.info/lrec2026-ws-nakbanlp-42
DOI:
10.63317/44f6f4a6cxma
Bibkey:
Cite (ACL):
Ali Adel Ali, Mona Khaled Ali, Mohamed Emad Sayed, and Ibrahim Naser Mostafa. 2026. Not Gemma at AR-MS NakbaNLP 2026: Mubsir OCR: End-to-End Recognition of Arabic Handwritten Text. In Proceedings of the 2nd International Workshop on Nakba Narratives as Language Resources @ LREC 2026, pages 273–278, Palma, Mallorca (Spain). ELRA Language Resources Association (ELRA).
Cite (Informal):
Not Gemma at AR-MS NakbaNLP 2026: Mubsir OCR: End-to-End Recognition of Arabic Handwritten Text (Ali et al., NakbaNLP 2026)
Copy Citation: