Constructing hierarchies for incomplete frames

Abstract

A frame is characterized as the hierarchical description with the object as the unit. On the other hand, it is not always true that the information is obtained by an interview with an expert or by observation in a systematic form. Then the obtained frames are described independently. It is desirable to organize the frames systematically by generating the class hierarchy. However, the frame contains incomplete information, such as undefined or indefinite attribute,and the traditional classification techniques based on the conceptual clustering cannot handle such information. This paper proposes a method to generate the class frame and construct the hierarchy for the frames with mixed undefined and indefinite attributes. In the classification of the frames containing incomplete information, the proposed method considers the attribute values other than the attributes describing the incomplete information and constructs the hierarchy by classifying the frames to the class containing the most similar frame. The validity of the generated classes, the effectiveness of the expansion procedure for the hierarchical tree, the validity of the evaluation criterion for the quality of the hierarchical tree, and the estimation performance for the attribute value in the classification of the undefined attribute are examined.