Enhancing Video Content Accessibility Through YouTube Transcript Summarization

In the rapidly evolving era of online content on YouTube, effective content discovery has become more challenging. This study addresses the objective of the problem of content discovery on YouTube by recognizing key issues, such as creating a summary with relevant content, generating subtitle-free video summaries, and lengthy video summarization. We aimed to fill these research gaps in this field. Our research innovates with a combination of three features that utilize state-of-the-art Natural Language Processing (NLP) techniques. Specifically, the summarization using advancements in the Bidirectional and Autoregressive Transformer (BART) model for the subtitled videos, integrates BART with Automatic Speech Recognition (ASR) for transcribing audio from non-captioned videos and implements a video clipping technique for handling lengthy videos effectively. Our primary goals are improving content summarization accuracy and enhancing user experience. We introduced Evaluation through Recall-Oriented Understudy for Gisting Evaluation (ROUGE) to achieve an average accuracy so that the user can check whether the model is generating a summary properly or not. In conclusion, this research shows advancements in YouTube video summarization, providing users with a concise, accurate summary for navigating YouTube videos.