Semantic analysis of spatial expressions in Japanese

There is continuing interest in geographic information science to explore people's spatial concepts and to formalize them to develop user-friendly geographic information systems (GIS). The cognitive linguistic approach appears to be effective to analyze spatial categories encoded in languages, to which this study is related. People perceive arrangements of entities in space, categorize them as spatial relationships, and describe them as spatial expressions in languages. Locative particles in natural languages, such as English prepositions “in,” “on,” “at,” etc., are used to describe skeletal structures of spatial configurations. To study categorization of spatial relationships encoded in languages is a foundation of GIS design, especially for spatial query languages, user interfaces, and reasoning systems.; Previous studies of the semantics of locative particles have mostly analyzed IndoEuropean languages, and more analyses on various languages, especially non-IndoEuropean languages, are needed to pursue the universals, as well as language-specific properties, of spatial conceptualizations embedded in languages. This study catalogues linguistic categories associated with three locative particles in Japanese, “ni,” “naka ni,” and “ue ni,” which roughly correspond to “at,” “in,” and “on” in English, by utilizing a conceptual model called the “use type model,” which is designed to model encoding and decoding processes in the mind. The model assumes that multiple meanings of a locative particle are derived from an ideal meaning and that people may have a set of constraints in the mind to decide which sense of the particle can be used in a given sentence to describe a specific spatial relationship.; Many use types identified in this study and their semantic components for the Japanese locative particles are relatively similar to the English counterparts proposed by Herskovits (1986), although some Japanese use types convey metric properties, which the English counterparts do not have. The use types seem to form prototype structures as English counterparts do. The set of the ideal meanings, however, has differences between Japanese and English. Regarding the usage constraints of the use types, there appears to be scale-dependency. In addition, possible interactions between the subject and the object of locative particles seem to be part of the usage constraints.