A Theoretical Grounding for Semantic Descriptions of Place

This paper is motivated by the problem of how to provide better access to ever enlarging collections of digital images. The paper opens by examining the concept of

place

in geographic theory and suggests how it might provide a way of providing keywords suitable for indexing of images. The paper then focuses on the specific challenge of how places are described in natural language, drawing on previous research from the literature that has looked at eliciting geographical concepts related to so-called basic levels. The authors describe their own approach to eliciting such terms. This employs techniques to mine a database of descriptions of images that document the geography of the United Kingdom and compares the results with those found in the literature. The most four most frequent basic levels encountered in the database, based on fifty possible terms derived from the literature, were road, hill, river and village. In co-occurrence experiments terms describing the elements and qualities of basic levels showed a qualitative accordance with expectations for example, terms describing element of beaches included shingle and sand and, those for beach qualities included being soft and deserted.