There are a lot of learning contents, that have the same contexts which are targeted specific devices, under circumstances in which diverse devices support a dasiaU-Learning environmentpsila. But the current methods selecting adequate contents have little flexibility. So, this paper we propose an analogy matched contents selection(AMCS) method which can find similar contents even though there are no exactly matched features of a learnerpsilas device. Also, our research shows that our technique is at most two times more efficient than existing techniques.