We extend the popular force-directed approach to network (or graph) layout to allow separation constraints, which enforce a minimum horizontal or vertical separation between selected pairs of nodes. This simple class of linear constraints is expressive enough to satisfy a wide variety of application-specific layout requirements, including: layout of directed graphs to better show flow; layout with non-overlapping node labels; and layout of graphs with grouped nodes (called clusters). In the stress majorization force-directed layout process, separation constraints can be treated as a quadratic programming problem. We give an incremental algorithm based on gradient projection for efficiently solving this problem. The algorithm is considerably faster than using generic constraint optimization techniques and is comparable in speed to unconstrained stress majorization. We demonstrate the utility of our technique with sample data from a number of practical applications including gene-activation networks, terrorist networks and visualization of high-dimensional data.

While zoomable user interfaces can improve the usability of applications by easing data access, a drawback is that some users tend to become lost after they have zoomed in. Previous studies indicate that this effect could be related to individual differences in spatial ability. To overcome such orientation problems, many desktop applications feature an additional overview window showing a miniature of the entire information space. Small devices, however, have a very limited screen real estate and incorporating an overview window often means pruning the size of the detail view considerably. Given this context, we report the results of a user study in which 24 participants solved search tasks by using two zoomable scatterplot applications on a PDA -one of the applications featured an overview, the other relied solely on the detail view. In contrast to similar studies for desktop applications, there was no significant difference in user preference between the interfaces. On the other hand, participants solved search tasks faster without the overview. This indicates that, on small screens, a larger detail view can outweigh the benefits gained from an overview window. Individual differences in spatial ability did not have a significant effect on task-completion times although results suggest that participants with higher spatial ability were slowed down by the overview more than low spatial-ability users.

Abstract

Maps are currently the most common application domain for ZUIs. Standard techniques for controlling such interfaces on pen-operated devices usually rely on sequential interaction, i.e. the users can either zoom or pan. A more advanced technique is speed-dependent automatic zooming (SDAZ), which combines rate-based panning and zooming into a single operation and thus enables concurrent interaction. Yet another navigation strategy is to allow for concurrent, but separate, zooming and panning. However, due to the limitations of stylus input, this feature requires the pen-operated device to be enhanced with additional input dimensions. We propose one unimanual approach based on pen pressure, and one bimanual approach in which users pan the view with the pen while manipulating the scale by tilting the device. In total, we developed four interfaces (standard, SDAZ, pressure, and tilting) and compared them in a usability study with 32 participants. The results show that SDAZ performed well for both simple speed tasks and more complex navigation scenarios, but that the coupled interaction led to much user frustration. In a preference vote, the participants strongly rejected the interface and stated that they found it difficult and irksome to control. This result enhances previous research, which in most cases found a high user preference for SDAZ, but focused solely on simple speed tasks. In contrast, the pressure and tilt interfaces were much appreciated, which, considering the novelty of these approaches, is highly encouraging. However, in solving the test tasks the participants took hardly any advantage of parallel interaction. For a map view of 600x600 pixels, this resulted in task-completion times comparable to those for the standard interface. For a smaller 300x300 pixels view, the standard interface was actually significantly faster than the two novel techniques. This ratio is also reflected in the preference votes. While for the larger 600x600 pixels view the tilt interface was the most popular, the standard interface was rated highest for the 300x300 pixels view. Hence, on a smaller display, precise interaction may have an increased impact on the interface usability. Overall, we believe that the alternative interaction techniques show great potential for further development. In particular, a redesign should encourage parallel interaction more strongly and also provide improved support for precise navigation.