5.2 Expert Modules of an IMMPS

  The knowledge present in an IMMPS may be located in a number of conceptually distinguishable knowledge bases which are difficult to explicitly characterize. However, starting from the experience of previous IMMPSs, some knowledge bases seem to be indispensable and therefore should be modeled as shared resources rather than as local or private resources. These are resources comprising knowledge of the application domain, the user, the context, and design tasks. They are reflected in the RM by expert modules of the Knowledge Server.

The knowledge present in an IMMPS may include, and often this is the case, explicitly encoded knowledge sources other than those listed in the table. For example, when designing a presentation system as a self-reflective system, one may add a knowledge source that comprises meta-level knowledge. However, for the purpose of this RM, we will classify a multimedia presentation system as an intelligent system, if it exploits at least the above mentioned knowledge sources to achieve presentation goals.

5.2.1 User Expert

The User Expert is the maintainer of the knowledge that forms the user model of an IMMPS. As suggested by, e.g., [12, 13], a user model may comprise representations of a user's:

• goals and plans;
• physical or mental abilities;
• attitudes and preferences;
• knowledge and beliefs.

With the current RM, however, we only intend to provide a conceptual location for such knowledge. Issues concerning completeness of user models, representation forms and reasoning processes are not addressed here. We restrict ourselves to mention that there are already multipurpose tools for building and maintaining user models (e.g., the user modeling shell system BGP-MS [14]), and also attempts towards standards for user modeling within that research community.

The clients of the User Expert are the layers of the presentation systems, and all the other experts since there are many decision points along the presentation generation process where the user model must be taken into account. For example, when selecting the content of a presentation according to the user's information needs, or when deciding which media to use so that a user's perceptual abilities and preferences are matched.

5.2.2 Application Expert

The Application Expert is to provide the components of the layers with application-specific knowledge. The following tasks are assumed to be carried out by this component:

• interfacing with the application system(s);
• information/data format conversion;
• provision of information/data;
• information/data characterization;

The first two tasks always occur when information has to be transferred from one system to another, say from an application to an IMMPS. By information/data provision we mean the process of making accessible the pool of information from which content can be selected. This task may involve reasoning processes and thus goes beyond pure information/data storing. Finally, we assume the Application Expert is able to perform a characterization of the incoming information/data.

5.2.3 Context Expert

The Context Expert serves as a container for knowledge concerning the context. It may be structured into:

• generation context; i.e., a representation of what has been generated so far. For example, the context expert may comprise a lookup table which contains the encoding relations holding between the media objects and their underlying sematic concepts.
• presentation context; i.e., a representation of what has been presented to the user so far, and, if applicable, in which way the user has interacted with (interactive) presentation parts.

The task of the Context Expert is similar to the role a discourse model plays in a natural language generation or a dialogue system. In particular, this component is responsible for the resolution of context-dependent references.

5.2.4 Design Expert

The Design Expert complements the other experts in the sense that it provides a container for all other knowledge which: (a) is relevant for decision making in an IMMPS, (b) does not fit into one of the above expert modules, and (c) should be modeled as a shared source as it will be accessed for solving different tasks. It may include:

• Media/Modality Model;
• Design Constraints;
• Device Model; a (partial) model of the computational environment, comprising for example knowledge about the characteristics and availability of the output(/input) devices.