Toyoaki Nishida. Kazuto Koujitani and Hideaki Takeda
An ability of organizing conceptually promiscuous data into a uniform structure is critical to knowledge navigating systems. In this paper, we describe a plain indexing method for accumulating diverse information originating from heterogeneous natural information sources. We present a couple of heuristic techniques: (a) orthogonal decomposition which allows for reorganizing a given information base into a Cartesian product of primitive concepts, and (b) analogical refinement which enables further elaboration based on the measurement of similarity. The proposal is implemented and evaluated against sample information bases containing about a few thousands conceptuai units. We discuss the strength and weakness of the method based on the analysis of experimental results.