Paper Reading #16: Performance Optimizations of Virtual Keyboards for Stroke-Based Text Entry on a Touch-Based Tabletop

Comments

Reference information
Authors: Jochen Rick
When/Where: UIST’10, October 3–6, 2010, New York, New York, USA.

Summary
In this paper, the author discusses a solution for more efficient text input on tabletop surfaces. Traditional, physical QWERTY keyboards generally perform well because they are at a proper distance and allow users to use all ten fingers. On tabletop surfaces, where virtual applications or environments can be used, these physical keyboards are awkward and reduce interaction space, so larger and virtual keyboards are more viable; however, the tabletop interface lacks tactile feedback, rendering the conventional "tap-based input" somewhat awkward and uncertain. Thus, an alternative is "stroke-based input", where users trace their fingers through the letters of the word in a single gesture. To further improve stroke-based input, the author also seeks to modify the key layout to better suit the input paradigm.

The author explores previous keyboard layouts and conducted a user study to examine how people would complete stroke sequences. From this, a mathematical model was created using Fitts's Law. Adjustments were made to existing keyboard layouts in order to evaluate their usage. The author found that a layout optimized for one purpose might be very problematic when applied to a different purpose. From the data retrieved by evaluating each keyboard layout, a new and improved layout was designed using a simulated annealing process with hill climbing.

The author concludes that more efficient keyboard layouts are possible, but it is difficult for people to adopt new formats. In addition, their speed gains (EX: Dvorak only increases efficiency by about 10%) are not great enough to warrant switching. But with the emergence of new technologies, such as handheld devices, combining different keyboard layouts with stroke-based input can yield gains in efficiency great enough to be worthwhile.

Discussion
I found the explanation of the creation of a model for stroke movement to be difficult, as it was rather technical. Having said that, I enjoyed the rest of the paper as well as the problem it addressed. I've known about the Dvorak layout for sometime now, but have never really wanted to try it because of the adjustment time required and because I have not wanted to look for or purchase a Dvorak physical keyboard. But as the author points out, different interfaces on handheld devices or tabletops, where ten-finger input is not intended, may be abstracted enough for me and others to be willing to adopt new layouts. I also liked reading about the history of keyboard layouts, as I didn't realize there were so many developed before, other than QWERTY and Dvorak.