Information Technology Reference
In-Depth Information
Chapter 9
Multimodal Output Management
Each time the dialogue manager decides to generate a message for the
user, which usually happens a short time after the end of the user's
intervention (but can also happen in the middle of an utterance), an automatic
generation process is launched. For written or oral dialogue, it is the domain
of text generation that is concerned. For multimodal dialogue, whether it is an
information system able to display complex data, a system managing a
microworld represented on a screen, a system with a force-feedback device,
an ECA or a robot able to generate gestures when talking, the generation of a
natural language utterance goes hand in hand with that of a gesture or a visual
return. The process can then involve multimodal generation, that is the
generation of multimodal references in the opposite direction as that we have
studied in Chapter 6, as well as the transmission of multimedia information.
For this last point, the field concerned is that of multimedia information
presentation systems, called IMMPS (Intelligent Multi Media Presentation
Systems) [STO 05], a full-fledged field of research similar to that of ECA. An
MMD system's output management can involve many processes spread over
multiple modules.
To approach these processes, we can draw a distinction between the what
and the how. The first falls within the scope of the dialogue manager
[JUR 09]. It integrates a what to say and potentially a what to display and a
what to do, each of them including semantic content and, especially in the
first case, a dialogue act. The second falls within the scope of text generation
Search WWH ::




Custom Search