Generative AI and Sentiment Analysis: An Isep PhD Advances Natural Language Processing
Transforming language understanding models through hybrid artificial intelligence — that’s the challenge successfully met by Souha Al Katat in her thesis defended at Isep. By combining transformers and knowledge graphs, she paves the way for a new generation of sentiment analysis that is more nuanced, more inclusive, and more responsible.
Sentiment analysis in text is a key challenge for applications in marketing, politics, healthcare, and public opinion. But how can we accurately capture nuance, context, and implicit emotions—especially in underrepresented languages like Arabic?
This is the question addressed by Souha Al Katat, a PhD candidate at Isep’s LISITE laboratory, during the defense of her thesis on July 3, 2025.
A Hybrid AI + Graph-Based Model for Better Language Understanding
Her research presents a significant advancement in aspect-based sentiment analysis (ABSA) through the development of a hybrid model called TrNvKG, which combines two powerful approaches:
-
Transformers — generative models at the core of modern AI, capable of learning complex language structures
-
Knowledge graphs — which enrich the model with contextual knowledge, enabling more accurate and nuanced interpretation
This hybrid approach makes it possible to:
-
Improve the detection of implicit sentiments
-
Better handle ambiguous contexts
-
Enhance the interpretability of results
-
Address the Arabic language, often overlooked in traditional AI approaches
Towards More Sustainable and Responsible AI
Beyond performance, Souha’s thesis also examines the environmental impact of these complex architectures. She offers concrete pathways for designing more energy-efficient and responsible artificial intelligence models, in alignment with the Sustainable Development Goals (SDGs).
International Supervision and a High-Level Jury
This research was supervised by:
-
Prof. Lionel Trojman – Isep
-
Dr. Hussein Hazimeh – Lebanese University
and was defended before an international jury composed of experts in natural language processing and AI:
-
Dr. Célia Da Costa Pereira – Université Côte d’Azur
-
Prof. Yudith Cardinale – Universidad Internacional de Valencia
-
Dr. Rafael Angarits – LIP6, Sorbonne University / Paris Nanterre
-
Dr. Ibrahim El Bitar – Lebanese University
-
Prof. Salima Benbernou – Université Paris Cité
-
Dr. Chamseddine Zaki – American University of the Middle East
-
Hussein Hazimeh – Lebanese University
A Thesis at the Crossroads of AI, Language, and Ethics
👏 Congratulations to Souha Al Katat for this major contribution that brings together artificial intelligence, natural language processing, linguistics, and digital sustainability. A breakthrough that reflects ISEP’s ability to support high-impact research projects with strong scientific and societal value.
