Comparison Between KLM Estimations and Screen Reader Times in Non-Visual Interaction

In this paper we present the comparison between the results of two measurements of non-visual performance: Keystroke-Level Model (KLM) estimation and screen reader times. The estimation is done on an online multiple-question exam using three non-visual interaction scenarios. In both measurements we consider different speech rates, since they are determinant of the reading performance. Results indicate higher times in the KLM estimation for most of the scenarios and speech rates, which suggests that improvement of the model is needed in order to characterize non-visual performance. On the other hand, results reinforce the need for good accessibility of Web contents in order to improve the efficiency of non-visual access.