Hybrid systems integrate different, complementary representations, each dealing with particular aspects of the domain. Combining different, complementary representations, as discussed in [78, pages 29--30,], is therefore a key issue (a multi-paradigm approach, from the point of view of the computational mechanism).
For example, taxonomic representations in terms of semantic networks are appropriate for inheritance and classification inference mechanisms [39]; production rules are appropriate for representing logical implications; reasoning on actions and plans requires still further mechanisms, such as temporal reasoning; finally, reasoning on the domain by means of ``metaphors'' is crucial for integrated multimedia representations.
In this framework, integrated agent architectures are promising paradigms
for the development of hybrid systems for multimodal integration and
multimedia representation [288,101,201]. Intelligent agencies
are shortly reviewed later on in this chapter, see 5.2
Esprit Project 8579/MIAMI (Schomaker et al., '95)