Aram Lab

The Arabic AI Modeling (Aram) Lab is a research lab at the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) led by Bashar Alhafni. Our mission is advancing research and education in Arabic natural language processing and machine learning, with an emphasis on developing linguistically informed and socially grounded AI systems.

Here are some of the research themes we are currently interested in:

Educational Arabic NLP

Developing Arabic NLP technologies for education, with a focus on both assistive and assessment applications. We currently focus on:

• Grammatical error detection and correction
• Pedagogically aligned Arabic writing assistants
• Automated essay scoring
• Arabic text simplification and readability assessment

Human-Centered Arabic NLP

Designing methods for human-centered, user-aware Arabic NLP across dialects, scripts, and styles. Our work spans:

• Dialectal Arabic text normalization
• Personalized generation grounded in Arabic linguistic traits and sociolinguistic variations
• Modern Standard Arabic–Dialectal Arabic machine translation

Arabic Linguistic Representation and Model Behavior

Investigating the Arabic linguistic competence of various models across morphology, syntax, and orthography. Our ongoing research explores:

• Morphosyntactic evaluation of Arabic and multilingual models
• The effect of diacritization on comprehension, reasoning, and text generation
• How tokenization design choices shape model behavior for Arabic