Information Technology Reference
In-Depth Information
particular, we did not ask the participants of our user study to optimize any pre-specified
quality metric. Indeed, ourlong-term goals include discovery of users' metrics.
Our new technique employs animation techniques found in modern computers games,
and many applications apply animation techniques to enrich the look and feel of the
user interface. Animations to the interface smooth the roughedges and sudden transi-
tions common in many current graphical interfaces, and strengthen the illusion of direct
manipulation that many interfaces strive to present [16]. Animation improves a user's
understanding of the direct manipulation of the data by better portraying such concepts
as constraints, relationships, and connectivity. These are powerfulcues for the direct
manipulation of large graph structures.
Ball, North, and Bowman [1] evaluated interaction techiques for large display visu-
alisations. They tracked physical navigation in 3D space via the participants head with
a VICON system. Their experiments found with increased size of the display, there was
more physical navigation. When combined with the reduced performance time on large
displays, they found a compelling suggestion that physical navigation was also more ef-
ficient. They also found that physical navigation was preferred over virtual navigation.
Peck, North, and Bowman [12] defined a new 3D interaction technique,
multiscale
interaction
, which associates the user's scale of perception to their scale of interaction.
Multiscale interactions exploit the user's physical navigation in front of a largedisplay
to directly control the scale of interaction, while adjusting their scale of perception.
Overall, they found evidence that multiscale interaction is a natural behavior, and this
technique can be useful in interaction designforlargehigh-resolution displays.
Skeletal animation
(see, for example, [8]) is a well established techniqueingraphics.
An object is modelled as a mesh, with “bones” as links between ”joints”. Movement of
the mesh in between keyframes can be computed using methods of
inverse kinemat-
ics
[17]. The skeletal animation techniqueismainlyused to animate people and ani-
mals, but Merrick et al. [9, 10] investigate application to graph interaction. Their work,
however, is limited to very small graphs.
3
GION: Graph Interaction Operation for Nodes
To untangle a graph
G
,theuser has to rearrange nodes by dragging them to new po-
sitions. Moving nodes one by one is time consuming. With thousands to millions of
nodes, the human resources required are too large both for untangling in practice and
for evaluation experiments like the one in this paper. Large graphs thus need interaction
methods for untangling that move more than one node at a time. The
GION
technique
uses ideas and off-the-shelf software from skeletal animation to simplify the process of
interacting with large graphs. More specifically, GION adapts a physics engine to move
many nodes at a time. GION treats the graph as a skeleton, where bones simulate edges,
and joints simulate nodes. However, the simple approach of representing every edgeas
a bone and each node as a joint does not scale to handle large graphs. Thuslarge graphs
are clustered to enable the user to move largenumbers of nodes at once. The physics
engine treats clustersasrigid bodies connected by joints. The effect of this approach is
that connected clusters move as a chain. This section describes the interaction technique
in detail and provides rationale for design choices.
