Information Technology Reference
In-Depth Information
Moreover, a dialogue is not only an exchange of information whether it is
explicit or implicit. It can also be a negotiation trying to convince the hearer
or to prove something. This aspect is not very prominent in MMD, but is the
focusofworkwhichleadstotheidentificationofspeechactswhichareinsome
way enriched by being put into perspective. Among the three types which we
have seen in Chapter 7, the assertion is, for example, not very explicit with
its underpinning intention. Baker [BAK 04] emphasizes that in a negotiation
dialogue, it is not about creating assertions as much as making proposals or
offers. An assertion can provide an argument, and this is also one of the aspects
of dialogue which is not greatly taken into account by MMD systems.
8.1.4. Dialogue, argumentation and coherence
Linguistics and pragmatics researchers such as J. Moeschler have long
studied the argumentative dialogue and drawn links with speech act theory
and even relevance theory. They thus define a specific category of speech acts,
the argumentation acts, that are carried out when an utterance is meant to
serve as a conclusion [MOE 85, p. 189]. The linguistic form itself being able
to contain argumentative instructions, for example argumentative connectors
“but”, “however”,“thus”,“because” or “so”, the interpretation of an utterance
gives it an argumentative dimension which consists of identifying the
(semantic) direction and type of (pragmatic) act.
As soon as several utterances are involved, it is also about linking them to
each other through argumentative relations which allows us to structure the
dialogue by detecting how an utterance can provide an argument in the
direction of such an assertion. All this gives an analysis dimension which we
have not yet mentioned and which consists of adding a set of additional labels
to each utterance, containing an indication on the act and the utterance's
direction and on its relations with previous and subsequent utterances. It is
actually here that we can observe that the implementations in MMD are very
few: this analysis dimension, a bit like that of the Geneva hierarchical model
of dialogue analysis, can mostly be applied afterwards on a corpus rather than
on an effective system in real time. Moreover, our ticket reservation examples
do not really involve an argumentative dimension, and it is mostly a challenge
for the recreational or open domain MMD systems.
Argumentative dialogue management is accurately described from a
theoretical point of view, especially by Moeschler [MOE 85], who suggests a
Search WWH ::
Custom Search