Description

Information and communication technologies (ICT) occupy an increasingly important place in our daily lives (CREDOC, 2021). Similarly, for the general population faced with these information and communication technologies, people with sensorimotor disabilities want and can claim socio-professional integration, access to knowledge through school integration facilitated by the Internet, access to leisure activities and social participation (Folan et al., 2015) (Picard, 2007) (Angelo et al., 2007) (Cornes & Bochel, 1987). Consequently, information and communication technologies are essential for people with sensorimotor disabilities and restoring this communication is fundamental.

Regarding access to digital tools, people with sensorimotor disabilities have difficulty accessing the pointing device (moving the mouse cursor, but also the various clicks) and input text (accessing the standard keyboard). Many assistive devices exist depending on the disabilities and abilities of the people, but also on their life habits, environmental constraints and expected uses of the computer tool (LoPresti and Brienza, 2004) (DeVries et al., 1998) (Yu-Luen Chen et al., 2003) (Y.-L. Chen et al., 2003) (Biard et al., 2011).

However, for people with complex sensorimotor disabilities, when few functional movements are possible, scanning access on a visual keyboard is preferred. At this stage, the movement of the mouse cursor, mouse clicks and text input are grouped under the same interface (visual keyboard) controlled by one or two switches. The visual keyboard presents the letters, the arrows for the direction of the mouse cursor and the different clicks (Figure 1).

In aims to increase people's text input speed, several studies have shown that the addition of linguistic prediction modules (of words or characters) would, in theory, make input text faster (Schadle & Poirier, 2004) (Trnka et al., 2007):

Word prediction (or completion) enables words or completions to be proposed in order to reduce the number of characters to be entered. By reducing the entry of certains characters, the text input speed should theoretically be increased. In fact, this is not necessarily the case (Raynal & Badr, 2022) (Pouplin et al., 2014).

Character prediction rearranges the layout of characters after each input, according to the probability of each one being selected. In this way, the most likely character becomes the one that can be accessed most quickly, with the least scanning. It has been shown that in the case of a pointing keyboard, this prediction does not satisfy users any more than word prediction (MacKenzie, 2008). Nevertheless, for scanning keyboards, which we are interested in here, even though there are few studies with real users, we can see that in some cases, character prediction seems interesting (Pouplin et al., 2014). Nevertheless, the main drawback of character prediction is that it results to a dynamic display of the keyboard. Indeed, the integration of character prediction seems to take place in much the same way from one study to the next, meaning a reorganization of the character set and therefore of the entire display. This prevents any memorization of the character layout for the user, and thus a certain amount of learning.

Therefore, we propose to integrate character prediction while keeping the display as static as possible.