Blog - Paper Reading #7

Performance optimizations of Virtual Keyboards for Stroke-Based Text Entry on a Touch-Based Tabletop

Authored by Jochen Rick.

Rick is a faculty member at Saarland University. He received his PhD from the Georgia Institute of Technology, and has done work utilizing wikis.

Summary

Hypothesis
This paper predicted that the type of keyboard layout would substantially alter user input speeds when using stroke-based text entry.

Methods
The experiment was simple. Each participant had to trace through a series of four points. He measured time taken, angle and distance between strokes and utilized this data to design a pair of keyboards. These were then tested against existing keyboard layouts for effectiveness.

Results
The optimized keyboards outperformed the existing layouts as predicted. He observed that Qwerty, Dvorak, and Wide Alphabetic performed badly. This was expected as these layouts are designed specifically for ten finger typing.

Contents
This paper opens with a short history on the origins of current keyboard layouts, describing the motivation and inspiration for the dominant designs. He then describes a short test performed to calculate efficient motion when utilizing stroke-based text input. Using his findings, he develops a mathematical model implementing simulated annealing to design an optimized keyboard. He then runs a series of efficiency tests on a wide range of keyboards and compares the result to the optimized versions.

Discussion
The findings seem convincing, though I have qualms with the testing procedure. Once the optimized keyboards were designed, they were not tested in human studies. The existing layouts were simply compared to the optimized layouts using the altered mathematical model. It seems logical that the keyboard designed by said model would outperform others in a test graded by the same model. Additionally, as the author observes, when typing on a handheld device, users tend to use two fingers. This study does nothing to account for that factor.
The importance of the work is clear. In this modern day, and increasingly large portion of the population owns a handheld device requiring text input. Today's keyboards are based on layouts that were designed with a different scale in mind. This miniaturization requires that our styles of input be reformed with other constraints in mind.
I believe that now Jochen Rick has his theory worked out, he needs to proceed to more rigorous testing. He is aware of the reason that the Qwerty keypad is favored in handheld devices, familiarity. What needs to be tested now is how easily can the optimized keyboard be learned, and what real advantages does stroke based text input offer over two finger typing.