Traditionally, the presentation order of the learning objects in a course must be described previously and manually. In a personalized tutoring system that may give different courses to different learners, planning the presentation order of courses can be irritating and time-wasting. This paper proposes an approach that can automatically composite and recommend courses to learners with different presentation order and in accord with their intentions. First, a course MAP is constructed according to the contents of related domain ontology and web pages collected form the Internet. Then, a learner's intention is analyzed for compositing automatically suitable orders of the learning objects to form a personalized course. This proposed approach can also recommend learning objects according to a learner's preferences and others' feedbacks. By this approach, personalized courses can be achieved more easily.