Envisioning a Hybrid Recommendation System for Filtering the Content from Short Videos

Short videos have become increasingly popular in recent years due to the rise of social media platforms. Many creators have moved away from making long form content and towards creating 6 to 15 second clips that feature humorous, viral, or relatable content. As well, people are more likely to watch short videos due to the limited amount of time for which they can devote their attention. This has become especially visible during the pandemic, as the amount of video consumption on social platforms has skyrocketed. In general, short videos offer an entertainment that requires little effort to view and consume. For a perfect short video, perfect content is purely necessary, in an analogous way as food is to humans. Platforms like YouTube, Netflix, Amazon, etc. use their own video recommendation systems for filtering out the contents and recommending the best accurate video based upon the user's choice. Filtering techniques such as content filtering, collaborative filtering, and deep learning-based filtering play important roles in the development of a good recommendation system. Most of the modern recommendation systems are also working upon the principle of ‘Hybrid Filtering’, wherein both content as well as collaborative filtering techniques are simultaneously used. In this study, use case of one of the popular online streaming platforms: YouTube is considered, which uses neural network-based collaborative filtering. As proposed research is upon the analysis of short videos, YouTube Shorts service are considered. Using the experimentation via machine learning algorithms, a comparative study between the three types of recommendation systems is developed. This research is concluded by jotting down the advantages of content filtering-based system over the collaborative one, and therefore suggest a hybrid recommendation system model which can perfectly filter out the contents of short videos being uploaded onto the YouTube platform.