Mona Diab
Institute Director, Full Professor, ACL Fellow, Language Technologies Institute
Prof. Mona T. Diab is the Director of the Language Technologies Institute (LTI) at Carnegie Mellon University and a Full Professor in the School of Computer Science. Dr. Diab is a globally recognized expert in computational linguistics and natural language processing (NLP). Her career spans academia, industry, and international leadership roles, including a tenured full professorship at George Washington University, Amazon AWS, and Meta, where she served as Lead Responsible AI Research Scientist and Technical Lead. An ACL Fellow, she has received numerous accolades, including the King Salman Global Arabic Academy award for her contributions to Arabic language technologies and recognition among the Top Global AI Scientists of Arab Descent by MIT Technology Review. As of July 2025, Dr. Diab will be serving as an elected member of the CRA board of Directors.
Dr. Diab's research focuses on advancing trustworthy NLP and responsible AI systems, particularly in low-resource and multilingual contexts. Her work includes innovations in controllable natural language generation, cross-lingual processing, computational social science, and health analytics. She has spearheaded high-impact initiatives, such as democratizing access to scientific content through machine translation (ACL 60-60) and addressing bias and compliance in AI systems. A pioneer in Arabic NLP, Dr. Diab has developed foundational resources and tools that bridge linguistic, social, and computational disciplines. She currently directs the R3LIT Lab (pronounced "relit") at CMU.
As an active member of the international research community, Dr. Diab has co-founded initiatives like the *SEM Conference and the Computational Approaches to Linguistic Code-Switching workshops. She serves on editorial boards of leading journals and advisory committees for global AI governance. With a steadfast commitment to mentorship, she continues to shape the next generation of computational linguists while fostering innovation at the intersection of AI and societal impact.
Research Areas
- Trustworthy NLP and Responsible AI
- Controllable Natural Language Generation & Conversational AI
- Culturally aware generative AI modeling and Evaluation
- Generative AI & Applied machine learning techniques
- Cross lingual/multilingual processing especially for low resource scenarios
- Evaluation & Annotation Science
- Computational social science/lexical semantics/sociolinguistics/pragmatics
- Social media/ (mental) health analytics and modeling
- Resource building, text analytics, information extraction, sentiment and emotion analysis,
- Arabic NLP/computational linguistics
- Democratization and Demystification of Scientific Content
Selected Publications
“Investigating Cultural Alignment of Large Language Model” (2024)
- Authors: Badr AlKhamissi, Muhammad ElNokrashy, Mai AlKhamissi, Mona Diab
- Summary: This study shows that Large Language Models (LLMs) align better with cultural knowledge when prompted in a culture’s dominant language and pretrained on culturally relevant data, introduces Anthropological Prompting to enhance alignment, and highlights the need for more balanced multilingual datasets to accurately reflect global cultural diversity.
“Automatic Generation of Model and Data Cards: A Step Towards Responsible AI” (2024)
- Authors: Jiarui Liu, Wenkai Li, Zhijing Jin, Mona Diab
- Summary: This work addresses gaps in AI documentation by introducing CARDBENCH, a large dataset of model and data cards, and CARDGEN, an automated LLM-based pipeline that improves the completeness, objectivity, and faithfulness of model and data cards to advance responsible AI practices.
“Semantic Word and Sentence Embeddings Compression using Discrete Wavelet Transform” (2024)
- Authors: Rana Aref Salama, Abdou Youssef, Mona Diab
- Summary: This paper demonstrates that applying Discrete Wavelet Transforms (DWT) to word and sentence embeddings can significantly compress embeddings—reducing dimensionality by 50–93%—while preserving or improving performance on semantic similarity and downstream NLP tasks.
"ALERT: Adapt Language Models to Reasoning Tasks" (2022)
- Authors: Pengfei Yu, Tianyi Wang, Olga Golovneva, Badr AlKhamissi, Saurav Verma, Zhenzhong Jin, Gokhan Gokturk, Mona Diab, Asli Celikyilmaz
- Summary: The study introduces ALERT, a methodology for adapting language models to enhance their performance on reasoning tasks.
"FEQA: A Question Answering Evaluation Framework for Faithfulness Assessment in Abstractive Summarization" (2020)
- Authors: Esin Durmus, He He, Mona Diab
- Summary: This paper presents FEQA, a framework designed to evaluate the faithfulness of abstractive summaries by leveraging question-answering techniques.
