Reference Resolution - Man-Machine Dialogue: Design and Challenges - page 105

terminals allowing several users to reserve tickets for the same trains
[KOL 10]. The point of a temporal model can also be found in a better use of
the verbs' linguistic characteristics: semantic classes, aspect classes
(inchoative or non-inchoative depending on the hypothetical beginning of the
action, terminative or non-terminative depending on its end), past participle
roles or even preposition roles. On certain points, the MMD system still has a
long way to go.
6.2.2. Analyzing the utterance “put that there”
The example “put that there”, which sparked the beginning of multimodal
dialogue [BOL 80], can be broken down in more detail when the following
figures are taken into account, all within the framework of speech-based
drawing software use:
- The reference “that” refers to a static object, which is part of a graphic
palette and cannot be moved. In this case, “put” refers to an action of
generating an identical copy and not a movement.
- The reference “that” refers to an object that is not stored in its correct
place (i.e. where all the other copies of the same class of objects are), or which
is not in the right configuration or direction. In that case, “put that there” might
be more than movement: it can also be a rotation or a sorting according to the
settings of the objects already sorted.
- The reference “that” depends on the result of the reference resolution
“there”: the location referred to by “there” might, for example, be a place
to sort objects of a certain category, and the accompanying gesture “that” is
potentially ambiguous between several objects of different categories.
- The reference “there” depends on the result of the reference resolution
“that”: it is the difference between “put carpet there” when pointing at an area
of the room, and “put a nail there” with the same gesture.
- The reference “there” depends on specific knowledge, for example when
“that” refers to an electric plug: the gesture accompanying “there”, even if it
is precisely carried out, cannot point at the exact location of the plug, since
it has to answer to height standards, or distance from another plug standards,
which take priority when placing it and lead to reinterpreting the gesture as an
approximate pointing gesture.