Improved video content ranking with analysis of transcripts and user feedback

Abstract

"Online material consumption has a significant increase by the day as people are working from home more, students are getting used to remote learning, entrepreneurs are trying to digitize their businesses so that they can operate from home. The pandemic has affected everyone’s life in some way and as a result, digitalization of anything and everything is on the rise.

Due to high demand of online material for day-to-day work, the supply of online content has also increased where employers, schools, universities, and individual content creators have been routed towards creating online material. Therefore, it is very difficult to find the most suitable and relevant content at a glance.

To solve this issue, video search results ranking should be benefited from the actual video content, constructiveness of user feedback and other metrics which allows the user to identify content-rich videos which has quality content, at a glance. The SeekReel system proposes an approach to re-rank video content based on transcript relevance and comment constructiveness in order to filter out non-content-rich videos from search results. The re- ranking proposed takes into consideration the caption relevance along with the context in comparison to the search term and combines it with a constructiveness score in order to obtain a final weighted average score for each video for the purpose of re-ranking.

With the use of this system, it is possible to rank informative, content-rich videos at the top of search results while de-prioritizing videos that got ranked because of other factors such as high number of views due to the content going viral. This would increase efficiency and productivity of people who rely on online content for education and professional work."