Grounding for a computational model of place

Abstract

Places are spatial locations that have been given meaning by human experience. The sense of a place is it's support for experiences and the emotional responses associated with them. This sense provides direction and focus for our daily lives. Physical maps and their electronic decedents deconstruct places into discrete data and require user interpretation to reconstruct the original sense of place. Is it possible to create maps that preserve this sense of place and successfully communicate it to the user? This thesis presents a model, and an application upon that model, that captures sense of place for translation, rather then requires the user to recreate it from disparate data. By grounding a human place-sense for machine interpretation, new presentations of space can be presented that more accurately mirror human cognitive conceptions. By using measures of semantic distance a user can observe the proximity of place not only in distance but also by context or association. Applications built upon this model can then construct representations that show places that are similar in feeling or reasonable destinations given the user's current location.

(cont.) To accomplish this, the model attempts to understand place in the context a human might by using commonsense reasoning to analyze textual descriptions of place, and implicit statements of support for the role of these places in natural activity. It produces a semantic description of a place in terms of human action and emotion. Representations built upon these descriptions can offer powerful changes in the cognitive processing of space.