2025 · 1st edition
AbjadNLP 2025
The inaugural workshop established a dedicated international venue for NLP across Abjad, Perso-Arabic and Ajami writing traditions.
AbjadNLP Workshop Series
Language technology for a script shared across continentsNatural Language Processing for Arabic, Perso-Arabic and Ajami writing traditions
AbjadNLP brings together researchers, language communities and technology builders working on languages written in Arabic-derived scripts. The workshop exists because script matters: it shapes tokenisation, orthography, OCR, search, generation, language resources and the ability of communities to participate fully in the digital world.
Why a workshop around writing systems?
Because language technology is not script-neutral. Orthographic variation, shaping and rendering, segmentation, transliteration, OCR, low-resource data and multilingual modelling all behave differently across writing traditions.
~1Bspeakers represented across the wider language communities
Why AbjadNLP matters
Digital inclusion starts with the languages people actually write
Arabic-derived scripts are used across Africa, the Middle East, Central Asia and South Asia by communities with very different linguistic histories. Some of these languages have mature digital resources; many others remain under-resourced, inconsistently encoded, poorly represented in training data or weakly supported by mainstream NLP systems.
That gap affects much more than benchmark scores. It affects whether people can search, translate, read, write, learn, access public information, build cultural archives and use modern AI systems in their own languages. It also affects whether linguistic heritage survives the transition from print and oral traditions into digital environments.
AbjadNLP provides a place to study these problems together while respecting the differences between Arabic, Perso-Arabic and Ajami traditions. It connects computational work with orthography, language resources, sociolinguistics and the practical needs of language communities.
The aim is not to force diverse languages into one technical mould. It is to build NLP that understands the diversity carried by a shared writing tradition.
01 · Representation
Many languages remain digitally under-resourced
Low-resource communities need corpora, lexicons, benchmarks, tools and models that reflect their actual writing practices.
02 · Writing systems
Script creates technical challenges of its own
Tokenisation, orthography, Unicode, glyph rendering, OCR, transliteration and spelling variation require dedicated attention.
03 · Modern AI
LLMs can reproduce existing resource inequalities
Strong performance in high-resource languages does not guarantee reliable behaviour for dialects, minority languages or alternative orthographies.
04 · Heritage
Language technology can help preserve written traditions
Better digital infrastructure enables communities to search, analyse, teach and preserve linguistic and cultural material.
Language coverage
A workshop defined by writing tradition, not by one language family
AbjadNLP deliberately spans languages that are typologically different but share Arabic-derived writing systems and many related computational challenges.
العربية
Arabic
Modern Standard Arabic, Classical Arabic and regional dialects, including challenges in morphology, dialect variation, code-switching, generation and evaluation.
فارسی · اردو
Perso-Arabic traditions
Persian, Urdu, Pashto, Sorani Kurdish, Sindhi, Uyghur, Azeri and other languages that use adapted Arabic-derived scripts.
عجمي
Ajami traditions
African languages including Hausa, Fula, Wolofal, Swahili, Kanuri, Mandingo and Tamazight written using Arabic-derived conventions.
What the workshop is trying to build
A sustained research community around inclusive Arabic-script NLP
AbjadNLP is designed as an ongoing workshop series rather than a one-off event, connecting research papers, shared tasks, proceedings and community-building across editions.
R
Resources
Encourage reusable corpora, dictionaries, datasets, annotation schemes, orthographic documentation and open tools.
M
Methods
Advance NLP and LLM methods that work across scripts, dialects and low-resource settings rather than only dominant languages.
E
Evaluation
Develop better benchmarks and shared tasks that expose where multilingual systems succeed, fail or behave unevenly.
C
Community
Connect researchers across regions and disciplines, including academia, industry and grassroots language communities.
Workshop series
From Abu Dhabi to Rabat to Athens
Each edition builds a permanent research record through peer-reviewed proceedings while extending the workshop's coverage of Arabic-script languages and shared research challenges.
2026 · 2nd edition
AbjadNLP 2026
The second edition expanded the community and introduced four shared tasks spanning generated-text detection, authorship, style transfer and medical NLP.
2027 · 3rd edition · Current
AbjadNLP 2027
The third edition continues the workshop's multilingual and community-driven agenda, with an expanded shared-task programme and wider Abjad and Ajami coverage.
A growing public research record
AbjadNLP proceedings are archived on the ACL Anthology, giving papers and shared-task work a permanent, openly accessible home and making it possible to follow the development of the field across editions.
Research themes
From foundational language resources to multilingual LLMs
The workshop welcomes work across the NLP pipeline, especially research that helps explain, document or overcome the technical consequences of writing-system and resource diversity.
Core NLP
Morphology, tokenisation, tagging, parsing, named entities, sentiment, language modelling and representation learning.
LLMs & Generative AI
Multilingual and low-resource LLMs, adaptation, evaluation, safety, generated-text detection and culturally aware generation.
Machine Translation & Speech
Translation, speech recognition, speech synthesis and cross-lingual technologies for under-represented languages.
OCR & Writing Systems
Optical character recognition, handwriting, Unicode, fonts, glyph rendering, transliteration and orthographic variation.
Language Resources
Corpora, dictionaries, benchmarks, annotation, orthography descriptions and tools for low-resource communities.
Culture, Society & Access
Code-switching, dialects, language preservation, education, cultural heritage, fairness and digital inclusion.
Workshop series
General Chair
ME
General Chair · AbjadNLP
Dr Mo El-Haj
Natural Language Processing researcher working on multilingual and under-resourced language technology, with a long-standing focus on Arabic NLP, language resources and cross-lingual research. AbjadNLP is organised as a continuing community workshop with collaborators and chairs changing across editions.