[Back]

J. Hahn, P. Fogliaroni, A. Frank, G. Navratil:
"A Computational Model for Context and Spatial Concepts";
Talk: 19th AGILE International Conference on Geographic Information Science, Helsinki, Finland; 2016-06-14 - 2016-06-17; in: "Proceedings", T. Sarjakoski, M. Santos, T. Sarjakoski (ed.); Springer, (2016), ISBN: 978-3-319-33782-1; 3 - 19.

A natural language interface will improve the human-computer interaction with Geographic Information Systems (GIS). A prerequisite for this is the mapping of natural language expressions onto spatial queries. Previous mapping approaches using, for example, fuzzy sets, failed because of the flexible and context-dependent use of spatial terms. Context changes the interpretation drastically. For example, the spatial relation "near" can be mapped onto distances ranging anywhere from kilometers to centimeters. We present a context-enriched semiotic triangle that allows us to distinguish between multiple interpretations. As formalization we introduce the notation of contextualized concepts that is tied to one context. One concept inherits multiple contextualized concepts such that multiple interpretations can be distinguished. The interpretation for one contextualized concept corresponds to the intention of the spatial term, and is used as input for a spatial query. To demonstrate our computational model, a next generation GIS is envisioned that maps the spatial relation "near" to spatial queries differently according to the influencing context.

A natural language interface will improve the human-computer interaction with Geographic Information Systems (GIS). A prerequisite for this is the mapping of natural language expressions onto spatial queries. Previous mapping approaches using, for example, fuzzy sets, failed because of the flexible and context-dependent use of spatial terms. Context changes the interpretation drastically. For example, the spatial relation "near" can be mapped onto distances ranging anywhere from kilometers to centimeters. We present a context-enriched semiotic triangle that allows us to distinguish between multiple interpretations. As formalization we introduce the notation of contextualized concepts that is tied to one context. One concept inherits multiple contextualized concepts such that multiple interpretations can be distinguished. The interpretation for one contextualized concept corresponds to the intention of the spatial term, and is used as input for a spatial query. To demonstrate our computational model, a next generation GIS is envisioned that maps the spatial relation "near" to spatial queries differently according to the influencing context.

Context, near, HCI, NLI, Spatial Concept