Multimodal Output Management - Man-Machine Dialogue: Design and Challenges - page 150

- semiotic constraints, linked to the emission of information for the
environment, with the speech rate quantity and quality, which, for example,
can turn out to be too intense for the communication environment.
Along this line, a third set of parameters covers the human aspects falling
within the scope of adapting to the user. It is first and foremost a question
of adapting to the user's physical abilities, in case of a disability, that is of
constraints on the communication channel operations, but also in general, with
theconstraintsandpreferenceswithregardtothelevelofuseofthesechannels:
if the audio channel is already in use, for example. It is then a question of
adapting to the user's roles and individual preferences. It is finally a question
ofhumanfactorsasuniversalpreferences, withphysiologicalfactors,linguistic
factorsandcognitivefactors.Thefirstarelinkedtothenatureofthemodalities:
in sound generation, such as a loud warning beep, the system has to know that
the higher pitchers are more strident than the lower ones, and that the louder
the warning the better the chances that it is perceived (but the more stressful
it becomes). For visual generation, the system can be led, as we saw in the
interpretation (section 6.1.2), to implement criteria from the Gestalt theory
perceptive grouping or even to take into account color theory observations, to
use the color red, that is the color that is most quickly perceived by a human
being, for urgent messages. Whatever the modality, the system can use notions
of salience and pregnance: a salient element is an element that stands out from
the others because of the unique properties, for example the only blue element
in a visual scene is easily found; a significant element, that is which has been
previously repeated to the point that the user's memory has recorded it, is also
more easily perceived.
The linguistic factors fall within the scope of the user's lexical, prosodic,
syntactic, semantic and pragmatic preferences. The system can adapt to these
preferences by aligning itself on the user's uses, that is using the same terms,
the same sentence structures, with a similar use of language. At the level of
natural dialogue in natural language, it is also a question of applying the
Gricean maxims when determining the message, minimizing the risks of
ambiguity and anticipating the ambiguity generation conditions (e.g. the
system can avoid a pronoun anaphora when there are several possible
antecedents), it can go so far as to avoid indirect and composite speech acts,
at least in simple dialogues where the task is prioritized above all. It is also a
question of using the information structure, especially to highlight part of the