Exploiting Reinforcement Learning to Profile Users and Personalize Web Pages

In this paper, we present a Web content adaptation system that is able to automatically adapt textual elements of Web pages, based on the user profile and preferences. The system employs Web intelligence to perform these automatic adaptations on single elements composing a Web page. In particular, a reinforcement learning algorithm, i.e. Q-learning, based on the idea of reward/punishment is utilized as the machine learning system that manages the user profile. Based on it, the user profile is updated, so that automatic adaptations can be effectively performed while surfing the Web. We created a simulation scenario to test our approach over different users with specific preferences and/or different kinds of disabilities. Simulation results confirm the viability of the proposal.