Dynamic composition components based on machine learning: architecture design and process

Abstract

The dynamic composition of components is an emerging concept that aims to allow a new application to be constructed based on a user’s request. Three main ingredients must be used to achieve the dynamic composition of components: goal, scenario, and context-awareness. These three ingredients must be completed by artificial intelligence (AI) techniques that help process discovery and storage. This paper presents framework architecture for the dynamic composition of components that can extract expressed goals, deduce implicit ones using AI. The goal will be combined with pertinent contextual data, to compose the relevant components that meet the real requirements of the user. The core element of our proposed architecture is the composer component that (i) negotiate user goal, (ii) load the associated scenarios and choose the most suitable one based on user goal and profile, (iii) get binding information of scenario’s actions, (iv) compose the loaded actions, and (v) store the new component as a tree of actions enabled by contextual or process constraint. In our e-learning proven of concept, we consider five components: composer component, reader component, formatter component, matcher component, and executor component. These five components stipulate that a course is the combination of existing/scrapped chapters that have been adapted to a user profile in terms of language, level of difficulty, and prerequisite. The founding result shows that AI is not only an element that enhances system performance in terms of timing response but a crucial ingredient that guides the dynamic composition of components.