Next: Interactions in Virtual
Up: Architectures and
Interaction
Previous: The cognitive coprocessor
In [141], Hill and his coauthors have defined five qualities for
multimodal and multimedia interfaces:
- Blended modalities: The user should be able to blend modes
at any time.
- Inclusion of ambiguity: Input modes that yield ambiguous or
probabilistic input are desirable when appropriate.
- Protocol of cooperation: Intervention on input and output
modules should be available at any time.
- Full access to the interaction history: The interpreted
levels of interaction history must be accessible on-line as well as after
the finishing of the interaction session.
- Evolution: The interfaces have to be open to improvements
without the need for a complete reimplementation.
By investigating traditional interface architectures, they found that
semantics and pragmatics are usually not shared across modalities. Another
drawback is the missing of ``felicitous interruption, information
volunteering, and intelligent displays'', which are considered to be
important interface building techniques. The different modalities of
traditional architectures are therefore ``noncommunicating components'', as
shown in figure 4.4 .
Figure 4.4
: The traditional structure of modality cooperation (Taken
from [141])
In the new structure proposed by Hill et al., some of the processing levels
have been unified across input modalities, whereas others have been
opened. Thus, ``applications [...] interact with a semantic
representation of user activities rather than a syntactic one.'' The
principle architecture according to this approach is shown in
figure 4.5 . It is based on the principle of uniform
access, which guarantees for a separation of interface aspects from
the application; and the compromise of almost homogeneous
representation, which tries to balance between performance (mainly:
speed) achieved by specialized device drivers and access functions on the
one hand, and homogeneity that allows the blending of modalities on the
other hand.
Figure 4.5
: A better structure for modality blending (Taken
from [141])
Next: Interactions in Virtual
Up: Architectures and
Interaction
Previous: The cognitive coprocessor
Esprit Project 8579/MIAMI (Schomaker et al., '95)
Thu May 18 16:00:17 MET DST 1995