Constraint-based graphical layout of multimodal presentations

Abstract

When developing advanced multimodal interfaces, combining the characteristics of different modalities such as natural language, graphics, animation, virtual realities, etc., the question of automatically designing the graphical layout of such presentations in an appropriate format becomes increasingly important. So, to communicate information to the user in an expressive and effective way, a knowledge-based layout component has to be integrated into the architecture of an intelligent presentation system. In order to achieve a coherent output, it must be able to reflect certain semantic and pragmatic relations specified by a presentation planner to arrange the visual appearance of a mixture of textual and graphic fragments delivered by mode-specific generators. In this paper we will illustrate by the example of LayLab, the layout manager of the multimodal presentation system WIP, how the complex positioning problem for multimodal information can be treated as a constraint satisfaction problem. The design of an aesthetically pleasing layout is characterized as a combination of a general search problem in a finite discrete search space and an optimization problem. Therefore, we have integrated two dedicated constraint solvers, an incremental hierarchy solver and a finite domain solver, in a layered constraint solver model CLAY, which is triggered from a common metalevel by rules and defaults. The underlying constraint language is able to encode graphical design knowledge expressed by semantic/pragmatic, geometrical/topological, and temporal relations. Furthermore, this mechanism allows one to prioritize the constraints as well as to handle constraint solving over finite domains. As graphical constraints frequently have only local effects, they are incrementally generated by the system on the fly. Ultimately, we will illustrate the functionality of LayLab by some snapshots of an example run.