AAG Annual Meeting 2003 - Spatialization Sessions

Session Title: Spatialization I
Time: Wednesday, March 5, 2003, 1:00 pm
Room: Esplanade B, 2nd Floor
Chair: André Skupin
Organizers: Sara Fabrikant and André Skupin
Specialty Group Sponsors: Cartography, GIS, Environmental Perception & Behavioral Geography

Guoray Cai, School of Information Sciences and Technology, The Pennsylvania State University, University Park, PA 16802. Email: cai@ist.psu.edu
Spatial Metaphors for Visualizing Similarities of Textual Documents

Spatial metaphors play important roles in conveying the similarity measures among documents and queries. This talk will summarize and compare spatial metaphors for digital libraries developed in both geographical information science and information retrieval research, spell out the relative advantages and disadvantages in visualizing the relevancy space. Choose the right metaphor for document retrieval can be difficult, due to lack of guidelines and theoretical foundations. The second goal is to link cognitive principles to the design of spatial metaphors for digital libraries. The potentials of combining multiple metaphors will be discussed.

Katy Börner, School of Library and Information Science, Indiana University, Bloomington, IN 47405. Email: katy@indiana.edu
Towards a Cartographic Map that Shows the Evolution of Knowledge Domains

Knowledge Domain Visualizations (KDVs) are a special kind of Information Visualization that exploit powerful human vision and spatial cognition to help humans mentally organize and electronically access and manage large, complex information spaces. Unlike scientific visualizations, KDVs are created from data that have no spatial reference, such as publications, patents, and grants stored in digital libraries.
Despite advances in visualization research and the application of visual perception and cognitive principles, KDVs are typically hard to use by non-expert users. This is due to the fact that the generated "data landscapes" are highly abstract.
However, most of us use cartographic maps quite efficiently, e.g., to find the location of the next conference, to determine the best travel route, to gain an overview what the city has the offer, to locate hotels near the conference venue, etc. Similarly, cartographic principles might help to find a knowledge domain of interest in a map that shows all currently existing knowledge domains, to determine how this knowledge domain of interest is connected to familiar knowledge domains, to identify its subareas, or major experts, and (review) papers that could be used as primary information sources.
This talk aims to identify cartographic techniques and design principles that can be applied to create static and dynamic maps of knowledge domains. It will demonstrate the utility and limitations of existing data mining and information visualization techniques for the generation of cartographic KDVs and discuss the theoretical problems, practical engineering issues, and usability issues that need to be resolved.

Sara I. Fabrikant, Department of Geography, University of California Santa Barbara, Santa Barbara, CA 93106, Email: sara@geog.ucsb.edu
Distance and Similarity in Semantic Networks

Most data reduction algorithms used in information visualization project relatedness in data content into straight-line metric distance, such that semantically similar documents are placed closer to one another in an information space than less similar ones. In own previous studies we have shown that people are indeed associating metric graphical inter-point distances with semantic similarity of text documents depicted in 2D. However, document similarity along links of a network override similarity judgments between documents based on straight-line Euclidean distance.
In this paper we investigate what kinds of proximity strategies viewers employ when conflicting notions of relatedness are jointly shown in a spatialized network display of Reuters news articles (e.g. metric vs. topologic proximity). We also investigate how cartographic symbolization principles (e.g. the use of visual variables, such as color or size) and differing instructions affect similarity judgments.
The construction of the tested semantic news wire space follows ontological modeling principles of generalization, association, and aggregation. Several display types were devised where content similarity between news stories is metaphorically mapped into (1) linkages between points (network-topologic proximity), (2) metric linkages between points (network-metric proximity), and (3) membership of links belonging to thematic regions (perceptual grouping proximity).
Empirical results from this and related studies can be utilized for deriving design guidelines to construct cognitively adequate spatializations for knowledge discovery in very large databases.

Session Title: Spatialization II
Time: Wednesday, March 5, 2003, 3:00 pm
Room: Esplanade B, 2nd Floor
Chair: Sara Fabrikant
Organizers: Sara Fabrikant and André Skupin
Specialty Group Sponsors: Cartography, GIS, Environmental Perception & Behavioral Geography

Marco Ruocco, Department of Geography, University of California at Santa Barbara, 93106 Santa Barbara, CA, Email: ruocco@geog.ucsb.edu
Landscape information, cognition and visualization: towards a theory of camera motion over landscape.

Our experience of landscape, conceptualized in terms of spatial knowledge, aesthetics and sense of place, is based in part on our perception and cognition of environmental information. Landscape visualization and animation can be used to access and manage environmental information to produce an enriched, enhanced and expanded experience of landscape, if suitable criteria of visualization design are adopted. This experimental study aims at verifying if variations in the trajectory of a virtual camera affect the viewers' experience of landscape. The findings lead to the development of the foundations of a cognitive theory of landscape camera motion, which can be applied in the context of documentary cinema, landscape architecture and electronic entertainment.

William Pike, Department of Geography, The Pennsylvania State University, University Park, PA 16802. Email: wpike@psu.edu
Concept visualization for web-based group collaboration

Information visualization techniques are commonly used to browse corpora of existing documents or datasets. Spatialization tools such as the Self-Organizing Map (SOM) assist analysts in post hoc discovery of concepts and concept relationships; in most fields of science, however, concepts are often defined collaboratively through dialogical interaction among experts. This study describes an SOM implementation in a web-based tool that merges information visualization with an online discussion medium based on the Delphi method. This e-Delphi system (http://hero.geog.psu.edu/eDelphi) enables discussants to interact with visual depictions of group thought as it develops. The SOM input data change as the course of discussion changes, enabling the mapped view of a group's concept space to evolve as the concepts themselves are explored through dialogue. Initially vague or ill-defined concepts may come to occupy larger or more well-bounded areas of the SOM. Some concepts are subsumed by broader topics, while others experience intermittent life on the map's fringes. Discussion participants may also explore the changing relationships among concepts; those that initially cluster in concept space may diverge, while previously disparate concepts draw nearer as their proponents find common ground. Our e-Delphi implementation also enables visualization of participant spaces, where neurons represent individual discussants. Here, map regions suggest factions of users with related opinions, and the spatial relationships among clusters indicate participants' intellectual similarity. This linkage of dynamic concept visualization and online discussion enables efficient, collaborative investigation of organizational belief, including important concept relationships and patterns of consensus or disagreement.

André Skupin, Department of Geography, University of New Orleans, New Orleans, LA 70148. Email: askupin@uno.edu
Going SOMwhere?

The Global Positioning System (GPS) offers the ability to determine positions on the earth's surface with great accuracy and at close spatial and temporal intervals. The emergence of capable handheld receivers has led to a growing ability to determine not just individual locations, but to map out spatio-temporal trajectories in the form of movement across space over time. This paper proposes to link those trajectories to the attributes of the encountered spatial features in order to develop a notion of "traveling through attribute space."
First, a spatialized representation of geographic features based on their attributes is derived through the Self-Organizing Map (SOM) method. Then, a spatio-temporal trajectory is projected on the SOM, leading to the visual emergence of an attribute space trajectory. This technique is implemented for a data set consisting of 3000+ U.S. counties and associated socio-economic attributes. The SOM geometry is combined with a number of GPS tracks captured mostly along the Interstate highways system. Visual results of the method will be shown, together with a discussion of possible applications, like the hands-on exploration of socio-economic patterns in urban areas.

March 5, 2003