Graph Based Abstractive Text Summarization of YouTube Comments

Vijayendra S. Gaikwad

Abstract: Summarizing documents or reviews has received a lot of attention in recent years due to the explosive expansion of online documents having large sizes and shopping websites having a large number of reviews. In this fast-paced environment, people require things to be done fast, saving time. Reviews and documents are being created in enormous quantities every day. Understanding the context of document/ reviews and converting them into a specific format, like a summary to conserve storage space and aid in information acquisition in a short period of time. We test a hybrid end-to-end model solution that synthesizes input video comments and abstract text summaries using natural language processing and graph-based methods. This methodology first provides an extractive text summary of the comments. We propose a joint end-to-end model using the FP-Growth method and T5 Model for generating an abstractive text summary of the video input provided.

Keywords: Machine learning, Deep learning, Natural Language Processing, Extractive & Abstractive Summary