A natural language processing based multipurpose summarizer system for online video transcripts and textual contents

Abstract

The present invention is related to a natural language processing based multipurpose summarizer system for online video transcripts and textual contents. The invention discloses a multipurpose summarizer that leverages transformer-based Natural Language Processing (NLP) models to generate concise and coherent summaries of both YouTube video transcripts and online textual content such as news articles and blogs. The system incorporates a data acquisition module for retrieving transcripts via APIs and extracting article text through web scraping, a preprocessing module with a context-preserving chunking algorithm to handle long-form inputs, and a summarization module employing abstractive models such as Pegasus and T5. The core innovation lies in the chunked summarization and integration mechanism, which processes large inputs within transformer token limits and merges segment-level summaries into a coherent, fluent final output. The invention further includes an evaluation framework that measures performance using ROUGE, BLEU, fluency, coherence, and compression ratio, ensuring summaries are accurate and human-readable. Summarized content is delivered in real time via a Flask-based web interface with options for export in TXT and PDF formats. The system significantly reduces information overload, saves time, and enhances accessibility, making it particularly valuable for education, media, research, corporate, and government applications.