Contextual video analytics and recommendations through natural language processing (NLP) and graph machine learning (GML)

This research paper explores an innovative approach to video recommender systems by integrating Natural Language Processing (NLP) techniques, graph databases, and advanced analytics. The smart content recommender system goes beyond traditional title or description-based recommendations, aiming to enhance user experience through content-level searching capabilities and spoken-word analysis within videos. Through sentiment analysis (SA), topic analysis (TA), and keyword extraction (KWE), the system tags and timestamps the video transcripts in a detailed and contextual manner and offers recommendations. Since the study is underpinned by Graph Machine Learning (GML) literature, it is informed by a call for recognizing keywords spoken in videos more efficiently. It responds to questions about recommending content with less explicit information and adds new solutions like voice search and multi-language interface to improve accessibility. This idea may appeal to industries that want to receive videos from the targeted population with state-of-the-art content analysis.