Exploring Deep Learning and Generative AI Techniques in Automatic Text Summarization

Abstract

The growth of regular digital content generation across various domains has made it necessary to develop an Automatic Text Summarization technique to logically narrow down the large volumes of information into concise summaries. Deep Learning has helped significantly to improve both extractive and abstractive summarization methods. This paper provides a review of Deep Learning and Generative AI techniques used for text summarization, covering most recent and key models and architectures, input representation methods, training strategies, datasets, and evaluation metrics. Even with many advancements, the need for high volume datasets, high computational demands, and many more challenges still exist. This review also discusses these challenges and proposes potential future directions, like few-shot learning and human-in-the-loop augmented evaluation metrics, to enhance the reliability and applicability of Automatic Text Summarization across diverse domains.