The authors describe a process for generating multimedia presentations from domain-dependent descriptions of relationships among media objects to be included in a presentation. The process is based on a relational grammar, which associates the underlying domain description with presentation aspects such as spatial layout, style aspects, and temporal layout. This means that the same underlying document can be presented with different style characteristics. Any one style description can be interpreted for multiple end-user platforms by, for example, displaying many objects simultaneously on a high-resolution screen or, alternatively, displaying fewer objects on a low-resolution screen and providing links to the others. Compelling examples illustrate the power of the approach.

The work is well motivated and sound. The details of the grammar may be difficult to grasp for those unfamiliar with the area, but this in no way detracts from the significance of the paper.