A Review on Reinforcement Learning based News Recommendation Systems and its challenges

摘要

Recommendation systems are helpful in both business perspective and user day to day life. These days online contents are generated in huge amount and due to this, users need a special recommendation application namely personalized News Recommendation, and it is highly challenging due to its dynamic nature. Therefore, getting a suitable and relevant news article for a user is difficult task. To address the above challenge Reinforcement Learning algorithms plays crucial role because these algorithms very much helpful in dealing with the dynamic environment and large space. This paper reviews the different Reinforcement algorithms namely Deep Q-learning network (DQN), Deep Deterministic Policy Gradient (DDPG) and Twin Delayed DDPG (TD3) to develop the news recommendation system and also mentioned the challenges faced by the reinforcement recommendation systems. In this study it was found that TD3 is best suited to develop the news recommendation system.