Recommender Systems for Personalized Applications

Users in different services are faced by huge amount of content, which are mostly irrelevant. Therefore, there should be recommender systems to filter more relevant contents. The model-based recommender systems try to learn users preferences and build user profiles based on their implicit and explicit interactions with the service. They use user profiles to show them more relevant content, but the highly relevant content is not the only goal of recommender systems. The filter bubble, i.e. intellectual isolation from diversified content, is caused by highly accurate recommender system which is not desired in many disciplines. Therefore recommender systems should be able to recommend relevant content while diversify the recommendations enough to bring the users out of their filter bubbles. The aim of this PhD is to design a recommender system to be used in content stream area which is accurate and is able to diversify the recommendations. This system should intelligently recommend news articles and streams to the users based on their profile and implicit and explicit feedback, to improve user satisfaction and optimize business owners revenue.