With the evolution of mobile devices such as laptops, tablets, smartphones, etc., the number of users has also increased, but it is difficult for some people due to their limited abilities. The Internet of Things (IoT) can make people's lives more convenient, and especially important for people with disabilities. This paper presents an approach to semantic representation using ontology. This approach is based on IoT to facilitate the real-time populating of the ontology in the context of adapting User Interfaces (UI) to people with disabilities such as sensory impairments (hearing and vision). IoT technology offers a great service to gain more autonomy and independence for people with disabilities.

Adaptive User Interfaces (AUI) have to exploit Artificial Intelligence powerful to adapt User Interfaces (UI)s to People With Disabilities (PWD). Thus, Machine Learning methods could master disabilities problems and barriers. In this paper, we propose a Reinforcement Learning (RL) based approach for resolving and surmounting PWD-UI interactions barriers since RL is good for learning good behavior. Hence, we have called the approach as RL-AUIAC as Reinforcement Learning of Adaptive User Interfaces for Accessibility Context. RL-AUIAC is based on three Knowledge Layers (KL)s depending on the kind of adaptation and the resolved problem at each layer. KL uses the Exploration-Exploitation dilemma to respond to the question: what to learn from each planned-adaptation sequences? In fact, Disability Knowledge Layer (DKL)learns UI structure behavior depending on the disability profile. Modality Knowledge Layer(MKL) learns facilities of adaptation on the basis of the couple. Platform Knowledge Layer (PKL) explores-exploits platform-knowledge to learn adaptation facilities on the basis of.

We apply the Model-Driven Engineering for adapting User Interfaces (UI) in order to improve their accessibility into interactive system (IS). The solution that we propose has to be generic to be applicable in any context and with any type of UI. Therefore, we have developed a model driven approach based on meta-model transformations. This approach generates modality-based UI models adapted to each kind of disability. Hence, we have developed an accessibility context described with the type of disability, the used platform, the user identity, his/her situation and the surrounded environment. This context is defined with an ontology which will be used in the adaptation process. To generate Platform Independent and Specific UI Models, we have proposed a set of metamodel transformation rules.

In order to ensure accessibility of user interface, it is highly recommended to elaborate a generic solution which is compliant to disable user requirements and best accommodates their particular needs. Therefore, we propose, in this paper, a generic MDA-based solution of adapting application's interface to accessibility context. It consists of generating, automatically, accessible adapted User Interfaces. Hence, we propose a multimodel approach by incorporating the modality of interaction into adaptation process. Consequently, based on MDA principals, we develop different meta-model transformations to provide an adapted User Interface model according to recent accessibility context information and a given non adapted User Interface.