A lot of heterogeneous data must be organized properly to be utilized efficiently, and hierarchical classification based on semantics of data is useful approach to organize data. The semantics of data for the classification attribute is often multiple. When data (objects) are classified to more than one child class, they are duplicated in the terminal classes, that requires the elimination of the duplications when the objects of a class are produced from the terminal classes. When the objects whose semantics is the semantics of a set of classes are required, we have to check that objects are not classified to the other terminal classes. This paper shows methods for these problems, where one of the objects in the child classes is selected to represent the object in the child classes.