With increasing access to the Internet and the wealth of material online, a Web-based self-teaching system has considerable educational value. Accordingly, we developed AEGIS (Automatic Exercise Generator based on the Intelligence of Students), which automatically generates questions whose difficulty level fits the achievement level of a student. However, it was implicitly assumed that all the questions were already categorized according to their subjects. In practice, this is not the case, but it is unreasonable (because of time and cost) to expect teachers to categorize each question into a suitable subject domain. Therefore, we need a method for categorizing questions automatically according to specified teaching concepts. This paper presents an automatic question categorization mechanism according to both a list of teaching concepts, called a Navigation List (NaviList for short), and the meaning of questions. We define an XML tag called a CONCEPT tag, which indicates a concept in a question, and an ontology, which is a hierarchical cluster of concepts. The method uses the tags and the ontology to categorize questions, based on the similarity between each category in a NaviList pre-composed by a teacher and an ontological concept specified by a CONCEPT tag in a question.