Individual and multipersonal fuzzy spatial relations acquired using human–machine interaction

An intelligent program was used that chose questions to ask of users so the fuzziness of the spatial relation concept would be reduced, suggesting that even with a relatively small, simple spatial database, significant semantic variation does exist.

Abstract

A major challenge in constructing intelligent geographic information systems (GIS) has been the specification of the meaning of spatial relations so they can be incorporated as linguistic terms in geographic query languages or as fuzzy relations in the geographic object model. This study used an intelligent program that chose questions to ask of users so the fuzziness of the spatial relation concept would be reduced. The users were required to provide only yes/no answers. From this question and answer session the program constructs a fuzzy base relation. Results obtained from human subjects are presented. Two geographic databases were used, each providing a different spatial context. For each database there were eight sessions: one session for each of the spatial relation terms, four terms for nearness and four terms for farness terms, Multiperson concept construction was explored using the agreement, global evidence, combined agreement and global evidence, and Zimmermann's method. Results suggest that even with a relatively small, simple spatial database, significant semantic variation does exist. This result has significant implications for maintaining semantic integrity within and between geographic databases as well as managing semantic variation.