Publications

A forthcoming chapter on diminutive formation and usage in North African varieties of Arabic, with attention to morphological patterns, dialectal variation, and the linguistic functions of diminutives across the region.

A roundtable discussion on how artificial intelligence is reshaping MENA politics research, with attention to regional data, methodological challenges, governance concerns, and the role of domain expertise in evaluating AI systems.

The study introduced the first comprehensive Arabic dataset for tracking missing words (ellipsis). Testing showed that while LLMs easily detect if a sentence has missing words, they severely struggle to locate and reconstruct those missing words.

Summary: This study combines appraisal theory and sentiment analysis to compare how judges, perpetrators, and YouTube viewers evaluate morality in high-profile courtroom discourse. It treats the audience as a third discursive actor and shows how public reaction extends the meaning-making of the trial beyond the courtroom.

Summary: AMWAL introduces the first domain-specific Arabic financial NER system, trained on 26K news articles and grounded in the FIBO ontology. It recognizes 20 entity types and reaches a 95.97 F1 score, outperforming CamelBERT and Wojood.

Summary: Using a 150,000-fatwa corpus from Islamweb.net, this study shifts attention from muftis to the spouses who ask the questions. It maps the lived concerns behind marital conflict and shows how online fatwa requests reveal recurring themes in spousal duties.

Summary placeholder: summary coming soon.

Summary: This paper presents the IUNADI system for the NADI 2023 shared task on country-level Arabic dialect identification in tweets. It addresses fine-grained dialect variation across Arabic-speaking regions by framing the problem as tweet-level classification and evaluating a competitive approach for robust dialect prediction in noisy social-media text.

Summary: IUEXIST applies multilingual transformer models to sexism detection on English and Spanish tweets in the EXIST shared task. The system compares a single XLM-R Large model against an ensemble with extended training data, showing that the ensemble improves hard cases while the single model remains stronger on softer labels.

Summary: This study analyzes Arabic tweets about COVID-19 using thematic coding, LIWC-based psycholinguistic profiling with a custom Arabic dictionary, and temporal discourse analysis to trace how public framing evolved during the early pandemic period.

Summary: Based on 992,593 English tweets from the first 14 weeks of the outbreak, this paper combines corpus linguistics, LIWC analysis, and conceptual metaphor analysis to show sustained analytical reasoning alongside persistently negative tone and shifting emotional patterns.

Summary placeholder: summary coming soon.

Summary placeholder: summary coming soon.