The Geography of Semantic Information Spaces

The use of spatial metaphors to depict complex database content has become popular within the information visualization community to deal with the rapidly growing volumes of data. These spatialized database depictions are known as information spaces, or information worlds. Spatialization not only allows to visually summarize and describe complex data collections, but also provides opportunities for visual query and sense-making of large bodies of information. Although information spaces are abundant and span a wide array of application areas mostly outside of geography (see for example Card et al., 1999) a structured approach based on solid theoretical foundations to formalize the underlying representational framework seems to be missing. More importantly, very little is known on the usability of such kinds of representations for knowledge discovery. Two concerns should be addressed from a usability standpoint: the use of spatial metaphors as a data reduction strategy, and the effectiveness of spatial depictions for knowledge extraction. As argued in this paper, an explicit and structured spatialization design strategy needs to be in place before usable information spaces can be constructed. The goal of this paper is to devise explicit design criteria that are generic enough to be context independent, but specific enough to represent the domain appropriately. Three design areas can be identified for this endeavor: the visual spatial structure employed to represent the world of information, the representation of meaning encapsulated in the database for knowledge discovery, and finally the potential experiential effects spatialized views have on information seekers exploring semantic spaces to satisfy a particular information need. This paper adopts three ontological perspectives to formalize the issues of information space design identified above. First, geometric reduction of database complexity (e.g. cartographic generalization) is formalized using basic geographic concepts, such as identity, location, magnitude and time. Here geometric generalization is understood as the transformation of high dimensional phenomena into lower dimensional geographic representations. Second, the concept of a Benediktine space is introduced to propose a structured approach for the semantic generalization. Benediktine spaces are semantic constructs that preserve properties of mapped entities and functional relationships between entities represented in an information space (Benedikt, 1991). Third, an information user-centered view of design is adopted. Geographic space is more than absolute space represented by Euclidean geometry or topology. Geographic space includes affordances, experiential properties and socially constructed meanings. The cognitive space approach focuses on how people explore information spaces and are able to decode meaning encapsulated in spatialized views for knowledge discovery. The power of the spatial metaphor is the key for connecting the three complementary spaces types (e.g. geographic, Benediktine and cognitive). Spatial metaphors not only facilitate the reduction of database complexity through geometric and semantic transformations, but they also act as sensemakers for information seekers to comprehend abstract data domains with the help of familiar geographic concepts. The outcome of this research serves two purposes. First, a solid spatialization design strategy for constructing usable information spaces will help improve communication between information seekers and database providers for efficient knowledge exchange. A solid framework for data generalization through semantic abstraction is a promising avenue to deal with the information overload problem, if data archives keep growing exponentially. Second, it is notable that most of the spatialization work is carried out outside of the GISciences, with the exception of a handful geographers, for example Couclelis (1998), Fabrikant (2000), Fabrikant and Buttenfield (1997, 1999), Kuhn and Blumenthal (1996), Skupin and Buttenfield, 1996, 1997), and Tilton and Andrews (1994). It seems obvious that GIScientists are well suited for this task. GIScience provides the perspectives of space and place, as well as the necessary visual, verbal, mathematical and cognitive approaches to construct geographic representations (National Research Council, 1997). A sound representational framework of space grounded on ontological and semantic principles can be transferred directly to the explicit geographic domain as a basis to reduce current limitations of how geographic space is represented within GISystems.