Abstract
We propose an approach for categorizing data by assessing the degree of relevance among them manually. We employ a visual interface that allows moving the data around in the display screen, with which a person can arrange the positions of data so that the distance between them indicates the relevance as how one perceives it. As this interaction iterating, categories emerge where the data concentrate geometrically. In this paper, we report the evaluation of the approach by prototyping a system.
