An evaluation of Dasher with a high-performance language model as a gaze communication method

Abstract

Dasher is a promising fast assistive gaze communication method. However, previous evaluations of Dasher have been inconclusive. Either the studies have been too short, involved too few participants, suffered from sampling bias, lacked a control condition, used an inappropriate language model, or a combination of the above. To rectify this, we report results from two new evaluations of Dasher carried out using a Tobii P10 assistive eye-tracker machine. We also present a method of modifying Dasher so that it can use a state-of-the-art long-span statistical language model. Our experimental results show that compared to a baseline eye-typing method, Dasher resulted in significantly faster entry rates (12.6 wpm versus 6.0 wpm in Experiment 1, and 14.2 wpm versus 7.0 wpm in Experiment 2). These faster entry rates were possible while maintaining error rates comparable to the baseline eye-typing method. Participants' perceived physical demand, mental demand, effort and frustration were all significantly lower for Dasher. Finally, participants significantly rated Dasher as being more likeable, requiring less concentration and being more fun.

Life in the Fast Lane: Effect of Language and Calibration Accuracy on the Speed of Text Entry by Gaze

“…Perhaps most importantly, even if the basic techniques are the same, they may differ significantly in important implementation details. This holds for the experiments with Dasher [11,14], and the same is true for soft keyboards. For instance, feedback modes have a significant effect on text entry speed [7].…”

“…Our participants were native speakers of Finnish and fluent with English. The participants in [11] were also fluent in English, but the paper does not tell if this means that they were natives. Participants having English as their mother tongue can be expected to get even better results than those in our study, even if the spelling controversies discussed in Section 5.2 still remain to some extent.…”

“…For [10], there were results for the learning phase (denoted by lp in Table 1) and the advanced phase (denoted by ap). In [11] there were two experiments, denoted by e1 and e2. Pedrosa et al [9] carried out a number of experiments; two longitudinal studies with able-bodied participants, followed by individual trials with users with motor disabilities.…”

“…The fact that the user is in control of adjusting the dwell time threshold during a session, as pioneered in [5], is of crucial importance. In [11], the dwell time was also adjustable, but the paper does not give details on how this was done. Judging from the screenshot of the interface in [11], this was only possible between the sessions.…”

“…In [11], the dwell time was also adjustable, but the paper does not give details on how this was done. Judging from the screenshot of the interface in [11], this was only possible between the sessions. It is understandable that users are not eager to jump to big decrements in the dwell threshold if they are unsure of how it affects them.…”