On personalizing Web content through reinforcement learning

Nowadays, a large use of personalization techniques is used to adapt Web content to users’ habits, mainly with the aim of offering appropriate products and services. This paper presents a system that uses personalization and adaptation techniques, in order to transcode or modify contents (e.g., adapt text fonts) so as to meet preferences and needs of elderly users and users with disabilities. Such an adaptation can have a positive effect, in particular for users with some reading-related disabilities (i.e., people with dyslexia, users with low vision, users with color blindness, elderly people.). To avoid issues arising from applying transformations to the whole content, the proposed system uses Web intelligence to perform automatic adaptations on single elements composing a Web page. The transformation is applied on the basis of a reinforcement learning algorithm which manages a user profile. The system is evaluated through simulations and a real assessment, where elderly users where asked to use the system prototype for a time period and to perform some specific tasks. Results of the qualitative evaluation confirm the feasibility of the proposal, showing its validity and the users’ appreciation.