Reusable System Architectures - Man-Machine Dialogue: Design and Challenges - page 69

development environment, or toolkit, will allow us to import the task's
specificities and to start the creation of the targeted system on this basis. We
then enter into the field of design-time architecture, that is the design of the
development environment rather than the MMD system itself.
4.2.1. Toolkits
Toolkits provide system developers with programming interfaces, i.e. the
technical abilities to implement the techniques used in MMD. There are other
kinds of toolkits that often integrate an architecture model as well as offer
techniques.Theyalsoofferaplatformtosupportthedevelopedcode'srunning.
More evolved, they can also emit recommendations based on the model they
implement, in relation with the software development standards.
McTear [MCT 04] highlights the use of toolkits. In general, most of the
major MMD research laboratories and computer science companies offer their
own development environment based on their experience in running systems.
Among other examples, we can mention the famous VoiceXML, the CSLU
toolkit or that of Carnegie Mellon, which are regularly taught in tutorials or
summer schools. The observation which is often made is that the available
toolkits help develop simple systems, but are far from allowing a programmer
to design a natural dialogue system in natural language. To achieve this goal,
all the NLP aspects have to be implemented in various languages, and that
falls more within the scope of NLP and MMD challenges. However, the tools
offered help develop systems and allow programmers to start designing an
MMD system without having to start from scratch as was the case in sections
3.1.1 and 3.1.2.
More recently, efforts undertaken in the field of model-driven engineering
(MDE) and in its subset that concerns architectures, i.e. model-driven
architectures (MDA), have reached the attention of man-machine
communication researchers, starting with the MMI specialists. These have
started reflections on the rebooting of the design cycle of an MMI so that
models and metamodels are taken into account during the very first design
phases, the last stages consisting of the automatic generation of MMI by the
development environment. Such reflections started off in MMD, and pursuing
them is a major challenge for the field. More specifically, the designer
specifies design models and model transformation processes (in a top-down