Amr Keleg
The university of Edinburgh
Edinburgh, Scotland
Hello (أهلًا وسهلًا)! 👋👋
My name is Amr Keleg عمرو قلج (/ʕamr/ /kɯˈɫɯtʃ/). I am a PhD student (CDT in NLP) at the University of Edinburgh, working under the supervision of Walid Magdy and Sharon Goldwater. I am currently studying the variation across and between the Arabic dialects, their mutual intelligibility, and the implications of this variation on the creation of multi-dialect Arabic datasets.
Feel free to ping me if you are interested in discussing ideas related to the following interests, and/or collaborating on that!
Research Interests
1) Computationally Handling Dialectal Variation (focusing on Arabic)
-
Estimating the Level of Dialectness Predicts Inter-annotator Agreement in Multi-dialect Arabic Datasets - ACL 2024 (Outstanding Paper award, Oral presentation)
Keleg, Amr, Magdy, Walid, and Goldwater, Sharon -
NADI 2024: The Fifth Nuanced Arabic Dialect Identification Shared Task - ArabicNLP 2024 (Shared Task Organization, co-located with ACL 2024)
Abdul-Mageed, Muhammad, Keleg, Amr, Elmadany, AbdelRahim, Zhang, Chiyu, Hamed, Injy, Magdy, Walid, Bouamor, Houda, and Habash, Nizar -
ALDi: Quantifying the Arabic Level of Dialectness of Text - EMNLP 2023
Keleg, Amr, Goldwater, Sharon, and Magdy, Walid -
Arabic Dialect Identification under Scrutiny: Limitations of Single-label Classification - ArabicNLP 2023 (Oral presentation, co-located with EMNLP 2023)
Keleg, Amr, and Magdy, Walid
2) Multilingual and Multicultural research
-
Revisiting Common Assumptions about Arabic Dialects in NLP - ACL 2025
Keleg, Amr, Goldwater, Sharon, and Magdy, Walid -
LLM Alignment for the Arabs: A Homogenous Culture or Diverse Ones? - C3NLP (co-located with NAACL 2025)
Keleg, Amr -
DLAMA: A Framework for Curating Culturally Diverse Facts for Probing the Knowledge of Pretrained Language Models - ACL 2023 (Findings)
Keleg, Amr, and Magdy, Walid -
An Unsupervised Method for Weighting Finite-state Morphological Analyzers - LREC 2020
Keleg, Amr, Tyers, Francis, Howell, Nick, and Pirinen, Tommi
3) Analyzing Romanized Writings of Non-Latin Languages
- I developed a rule-based tool transliterating Arabizi (Romanized Arabic) into Arabic script.
Other Interests
As an undergraduate student, I was a competitive programming addict (lots of fun experiences 😄). I am also an advocate of open-sourcing data/models/projects (twice a Google Summer of Code student for Apertium, and GNU Octave + contributor to other projects like Facebook/Duckling).
Note: The list of resources created as part of my research can be accessed through this page.
News
| Jul 14, 2025 | Our paper “Revisiting Common Assumptions about Arabic Dialects in NLP“ got accepted to ACL 2025! |
|---|---|
| May 8, 2025 | Gave a talk titled “Incorporating Sociolinguistic Theories for a Better Modeling of the Arabic Varieties” to the CardiffNLP. Check the slides: here. Thanks, Nedjma, for the invitation! |
| Mar 1, 2025 | My position paper “LLM Alignment for the Arabs: A Homogenous Culture or Diverse Ones” is accepted to the C3NLP workshop co-located with NAACL 2025! |