Content-based recommender system by user's visual attention

Nowadays, some of the world's biggest entertainment among the population is the consumption of streaming services such as movies or series through digital platforms such as Netflix, Disney+, Prime Video, HBO Max, Hulu, and Apple TV+. Traditional recommender algorithms process explicit information where the user indicates which movies he/she liked on the platform. However, implicit information like the user's visual attention adapted to personal interests may lead to better personalized recommendations. This has motivated the development and innovation of movie recommendation systems. Hence, we propose a novel recommendation algorithm based on the user's visual attention. The algorithm was implemented on a digital platform named Poppukōn where the movies are presented on the platform according to different recommendations like genre, ratings by likes and stars and two based on eye tracking features. The Poppukōn platform was implemented using the Django framework for the Back-End and the Vue framework for the Front-End. Experiments were carried out to find out the user satisfaction about the personalized recommendations presented on the platform. Our results provide evidence that using eye tracking features in a movie recommender system improve the user satisfaction using implicit (gaze) and explicit (rating) information.