Recent developments on mobile devices and wireless technologies enable new technical capabilities for the learning domain. Nowadays, learners are able to learn anywhere and at any time. The dynamic and continually changing learning setting in learner's mobile environment gives rise to many different learning contexts. The challenge in context-aware mobile learning is to develop an approach building the best learning content according to dynamic learning situations. This paper aims to develop an adaptive system based on the semantic modeling of the learning content and the learning context. The behavioral part of this approach is made up of rules and metaheuristics to optimize the combination of pieces of learning content according to learner's context.