A Comprehensive Narrative Review of Extractive Text Summarization: Techniques, Transformer Models and Open Challenges (2010–2026)
Baraa A. Elhady
*
Department of Information Systems, Faculty of Computers and Artificial Intelligence, Helwan University, Cairo, Egypt.
Osama E. Emam
Department of Information Systems, Faculty of Computers and Artificial Intelligence, Helwan University, Cairo, Egypt.
Helal A. Suleiman
Department of Information Systems, Faculty of Computers and Artificial Intelligence, Helwan University, Cairo, Egypt.
*Author to whom correspondence should be addressed.
Abstract
Automatic text summarization is critical for managing information overload in the digital era. This comprehensive narrative review examines extractive text summarization research published between 2010 and 2026, synthesizing classical statistical methods, machine learning approaches, and contemporary deep learning and transformer-based architectures. The review provides a unified taxonomy of extractive methods, comparative analysis of prominent models (including TextRank, BERT, RoBERTa, GPT, and T5), evaluation trends across major benchmark datasets and metrics, and consolidated evidence on persistent research gaps such as long-document handling, computational efficiency, cross-domain robustness, interpretability, and multilingual support. Findings indicate that transformer-based models achieve strong benchmark performance but still face practical limitations in scalability and context length, highlighting clear directions for future research and deployment-ready extractive summarization systems.
Keywords: Extractive summarization, natural language processing, comprehensive narrative review, transformer models, BERT, deep learning, evaluation metrics, literature review