Multimodal Output Management

Abstract

Intelligent multimedia presentation systems (IMMPS) have to take into account pragmatics and human factors such as the specificities of human perception, attention, memory, conceptualization, and language. Using a conversational animated agent or not, some principles can be followed to increase the communicative abilities of interactive systems. In this paper we propose a set of such principles. We exploit our background in natural language processing and computational pragmatics to provide specifications for multimodal systems. The classifications and principles and architectural concerns we present are based on some experimental observations (that are not described here) and constitute a kind of white paper for future implementations.