Semantics-Preserving Abstractive Summarization for Long-Form Text

The explosion of digital content has intensified the demand for automated summarization systems capable of transforming lengthy text into concise and meaningful representations. This work introduces Light Reads, an abstractive summarization framework developed by fine-tuning a T5 transformer model. Trained on a corpus of BBC news articles, the system applies modern Natural Language Processing (NLP) techniques to generate summaries that retain essential information while remaining succinct. To support practical use, Light Reads is equipped with an intuitive Gradio-based interface, enabling users to input text and obtain summaries instantly—making it suitable for applications across journalism, academia, and general information processing. The fine-tuned model shows strong performance, with notable improvements in ROUGE scores compared to baseline systems. This paper outlines the overall system design, fine-tuning methodology, and experimental evaluation, demonstrating that Light Reads produces coherent, information-rich summaries. We conclude by highlighting potential areas for expansion, including multilingual capabilities and enhanced handling of extended or complex documents.