Page 155 - Handbook of Electronic Assistive Technology

Chapter 5 • Alternative Access Technologies 143

Enhanced speed of text generation may also be achieved by abbreviation expan-
sion (AutoCorrect), where vocabulary is stored in the system under an abbreviation;
the user writes the abbreviation, which is then expanded by the access software. Again,
there is little published research to support the benefit of this technique. Abbreviation
expansion avoids the requirement to look at the list of predicted words displayed (which
is what seems to reduce text generation rate with word prediction) and so is likely to
improve the rate of text production. It does, however, also increase cognitive demands
and memory load.
Other rate enhancement techniques and strategies that are employed particularly
in speech-generating devices include (Beukelman and Mirenda, 2013): vocabulary and
selection set design to give fastest possible access to core vocabulary; phrase storage; and
use of multimeaning icons (i.e., Minspeak, as used in the Unity vocabulary from Prentke
Romich Company). Refer to the chapter on AAC for more information on these methods.

Speech Recognition

Speech recognition has several applications in terms of access to assistive technology:
control of devices and environmental control; personal communication for people with
dysarthric speech; as a writing tool for users with physical or literacy difficulties; and for
access and control of assistive technology.
Speech recognition has been researched and applied as an access method for elec-
tronic assistive technology for many years with varying degrees of success (Koester, 2001;
Hawley, 2002). Early speech recognition systems required considerable training and per-
sonalisation to recognise speech, the vocabulary that could be recognised was limited and
recognition accuracy was relatively poor (Juang and Rabiner, 2004).
Currently, speech recognition systems do not, for most users, require training to the
user’s voice; claimed accuracy rates are in excess of 90% (e.g., Microsoft claim an accuracy
of 94.9% for conversational speech (Xiong et al., 2017)) and as a result of this and of devel-
opments in natural language processing, speech recognition functionality is now available
on smartphones, tablets and computer operating systems, as well as dedicated ‘digital
assistants’ such as the Amazon Echo, Google Home and Apple HomePod. Speech recogni-
tion is now a free or low-cost mainstream technology that is available to all users. This is
likely to have significant implications for some users of electronic assistive technology.

Personal ‘Digital Assistants’

Software such as Siri for iOS and MacOS (Apple, 2017b), Google Now (Google, 2017b) and
Microsoft Cortana (Microsoft, 2017b) provide speech control for smartphones, tablets and
computers. Voice commands can be used to control a limited number of functions on the
device: for example, to search and interact with internet services, play music, create diary
events and reminders and launch applications by voice. These systems require an internet
connection because the recognition of speech into actions takes place on a remote server,
not on the user’s own device.