Doaa Sulaiman at AR-MS NakbaNLP 2026: Faithful Diplomatic Transcription of Arabic Manuscripts Using a Human-Centred Annotation Framework
Year: 2026
Type: conference-paper
Abstract: This paper describes participation in the Human Transcription Track (Subtask 1) of the NAKBA-NLP 2026 Arabic Manuscript Understanding Shared Task, which focuses on historical handwritten documents related to Palestinian Nakba narratives.The participant manually transcribed 500 cropped line images and designed a comprehensive transcription guideline from scratch, documenting all editorial decisions in a reusable framework.A faithful diplomatic transcription philosophy was adopted, preserving original spelling, punctuation, diacritics, numerals, and layout features without editorial normalisation, to create research-grade gold-standard data for downstream NLP and digital humanities research.Building on this philosophy, a 26-convention annotation framework was developed, organised into three layers: editorial-structural symbols, faithful-copying rules, and documentation labels, supported by a four-step quality-control pipeline.The submission achieved full coverage of all 500 assigned lines and attained an official CER of 0.02, WER of 0.01, and accuracy of 0.98.
Author: Doaa Bahjat Sulaiman
Institution: Istanbul University
Language: English
Subfield: Literature and Literary Theory
Field: Arts and Humanities
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W7166542339