Reusable System Architectures - Man-Machine Dialogue: Design and Challenges - page 72

during its running, update certain models on which it relies. If it has learning
abilities, it might want to enrich a model with new knowledge. If it has
confidence or relevance calculation scores, it might want to set the data of a
model by using these scores, so as to reinforce the impact of certain settings
compared to others. Theoretically, as the MMD system is increasingly used,
so it is susceptible to question the models on which it has been built. We
could then imagine a phase in which the MMD system decides to update itself
on its own by restarting all the derivation process.
4.3. Conclusion
A finalized dialogue system is designed to help the user carry out a given
task. Any system is thus optimized for a specific task and reveals itself to be
poorly performing for a different task. If we take into account the major effort
in design, the question arises of reusing the components from one system to
the next, and beyond, the question of the possibility of designing generic
components, which can be used regardless of the task since they would have
flexible settings. This chapter describes the question of genericness through
design and software architecture examples, with, on the one hand, the notion
of run-time architecture, to the system's running, and, on the other hand, the
notion of design-time architecture, with, for example, the model-driven
development, whose principle is to go from a set of models describing
resources and processes and automatically derive the modules that will
constitute the final system.