Section: Scientific Foundations
The purpose of multimedia document adaptation on the Web is to customize content for the variety of devices and networks that are now sharing the Web with traditional desktop computers. As a result of these changes, the Web infrastructure need to be reconsidered as a device-independent architecture, where information resources can be efficiently accessed with various types of devices and networks.
To reach this goal, the WAM project is following two complementary approaches: a comprehensive, global approach that requires several changes in the current Web infrastructure, and an approach that uses the current infrastructure and legacy content for adaptation.
There is no general solution to the problem of device independence today. Most efforts are rather dedicated to the development of good practices. To make progress towards a solution, two aspects are considered in the project: device-independent architectures and automatic content adaptation.
Ideally, content would be created or generated in a single universal format that could be delivered ``as is'' to any conceivable device. In practice this seems impossible, so the real techniques seek to minimize the number of variants needed, each variant being targeted at as wide a range of devices as possible.
The problem of adpative infrastructures is addressed through profiles, negotiation protocols and transcoding techniques. A profile is a formal representation of the context in which content is used: user abilities and preferences, device capabilities and limitations, network characteristics, etc. These descriptions must cover both static and dynamic parameters since the system conditions may change over the time.
To transform multimedia documents, one can rely on their semantics. The semantics considered here do not deal with the document content, but with the composition that is made explicit in a Web document:
Temporal semantics: in what order and when should each piece of information be presented to the user.
Spatial semantics: what are the relative positions of the document components on the display space.
Navigational semantics: how are pieces of information related in the hypertext network.
With this approach, adaptation can be done in very general semantic terms, independently from the multimedia objects. This makes it also possible to abstract (model) existing content into a unified representation, and then to facilitate the adaptation process.