Beyond Querying - User-Centered Data Management - page 35

boring, confusing, and time consuming activities. As an example, let us consider the editing of a very
long document (i.e., thousand of pages) that requires only scrolling along the y -axis. Without some
additional techniques, it becomes very hard to locate a specific page or to understand the part of the
document one is looking for. If we switch to a very large two-dimensional map, used by a user who
needs to explore only the details on a few parts of the map (e.g., the start street and the arrival street
of a long trip), the situation is even worse; it requires changing the zoom level during the interaction
and to scroll the image on both axes.
An effective solution for dealing with this problem is to arrange the presentation in such a way
that it always presents the user with an overview as well as the details in the representation, clearly
visualizing the relationships that exist among these two views. This technique is called either focus
plus context or overview plus detail . In Figure 3.15 , two examples are presented, one for browsing a
map (a) and another for a Word document (b).
A second strategy, which still allows for presenting the whole image in an effective way is
called distortion . The idea is to enlarge, in some way, only a part of the image, providing details in
that part and leaving the overall image available for providing an overview. In Figure 3.16 , we see
two examples of such a technique. In table lens ( Rao and Card , 1994 ), depicted in Figure 3.16 (a),
the representation is a simple Excel spreadsheet in which numerical values are coded through bars,
and it is possible to enlarge one or more rows to read the figures underlying the bars. The perspective
wall ( Mackinlay et al. , 1991 ) allows for browsing a large horizontal figure that is distorted on three
planes, getting details on the central part and preserving the overall perception of the image. In
Figure 3.16 (b), this technique is used to browse a file system where rows represent types of documents
and columns the creation date.
Semantic zoom is a technique that removes/resumes details from an image according to the
level of zoom. A very known example comes from Google Earth in which, for example, city names
appear only when the zoom factor is above a specific threshold. Semantic zoom can be usefully
combined with pan for preserving the overview. In Figure 3.17 , an example from Google Earth
is shown: while panning from New York to Rome, Google Earth performs a semantic zoom that
allows the end user to retain an overview of the whole transformation.
Problems may arise while visualizing large data sets, very likely producing cluttered images,
and they have been thoroughly analyzed by the Infovis community and a survey can be found
in ( Ellis and Dix , 2007 ). Decluttering techniques are classified in three groups: appearance, spatial
distortion, and temporal. Appearance includes techniques affecting the presentation look: point size,
altering opacity and color, and sampling and filtering. Spatial distortion encompasses point or line
displacement, topological distortion stretching the presentation, either non-uniformly (e.g., Fish-
eye) or uniformly (e.g., zoom), representing data items as single pixel, space filling, and dimensional
reordering. Animation techniques use the time axis to reduce clutter. As an example, Figure 3.18 (b)
shows the result of a non-uniform sampling technique ( Bertini and Santucci , 2006 ) able to rescue
density differences in a crowded Figure 3.18 (a) bidimensional scatterplot (postal parcels plotted by
weight and volume).