|
Speech Recognition for Smart Homes10. Usability criteria
It is sensible to develop smart home VI usability criteria. Users of current ASR systems may
well appreciate the frustrations of mis-tuned and under-performing systems where
backtracking and corrections require more effort and time than does the process of input
itself. Many give up, preferring to switch back to a clumsy, but reliable, keyboard.
Despite the theoretical reasons for adopting a VI system to free users from the constraint of
KMM, most users have in fact grown up with such constraints and are not unhappy with
them. By contrast, users of current ASR systems tend to experience mostly frustration in
their interactions - a major reason why LVCSR is not particularly popular today. It is
therefore only when viable alternatives to the KMM have been demonstrated in niche areas
that the general public will adopt a new perspective on the use of computer technology
without touch and vision-based user interfaces. Such a niche area is likely to be within the
smart home context, where significant advantages exist for ASR: a limited set of users,
relatively constant acoustic characteristics, constraints upon the tasks to be performed and
so on.
The assumptions for a smart home are that training is performed in advance, the major
sources of acoustic interference (such as infotainment and gaming systems) are under
control or at least linked in to the smart home electronics, the VI operational syntax is self-
contained, limited and known to the user, and that the user has had time to become familiar
with the system.
Major performance criteria include the percentage of tasks which are completed without
secondary user intervention (and then the degree of intervention required for those tasks
which are not completed straight off). For use by the general public, a useful performance
measure may well be how often the user must resort to a KMM solution for performing VI-
oriented tasks. However, as with all technology deployments to the general public, a final
successful verdict can only be pronounced when sales figures begin to indicate mass-
marked adoption of such technology.
|
| |
