Taleb Ben Diab ,Bashir2025-11-062025-11-062024-09-30https://dspace.univ-tlemcen.dz/handle/112/25194This thesis presents an advanced Arabic text summarization system built by fine tuning a T5-base transformer model on the AGS-Corpus, a curated Arabic summarization dataset. The model effectively handles the language’s rich morphology, diverse dialects, and complex syntax. Evaluation using BERTScore and ROUGE shows that the system generates high-quality summaries closely aligned with human references in content, fluency, and semantic similarity. To improve accessibility, the system is deployed as a user-friendly web application using Streamlit, enabling real-time summarization for non-technical users. Future work includes expanding the dataset to include more dialects and exploring alternative architectures like BART or GPT to further enhance performance, especially on complex or lengthy textsanglaisCreation of an automatic summary tool for arabic textsThesis