by Oscar Rosell
Cuypers is a system for generating tailored multimedia presentations for a wide diversity of devices and user preferences over the content. The system is divided in layers (of abstraction) from the rhetorical structure (which describes the relations that holds between the different media items) to the final presentation. In [1] there is a remark of the possibilities of incorporating some metadata in the output that explained which decisions had been taken in the process. In the refered document is stated that the metadata could be used, per example, for feedback and error detection. What I want to show is how all this metadata could be used for adapting the presentation client-side.
The problem with Server-Side tailoring, as in Cuypers, is that it's impossible to generate presentations that fit all the possible devices that could exist. In the Cuypers system itself, another problem is that any modification (little or not) in the client device that further constrains its capabilities, results in the impossibility of displaying the presentation as it is, and the necessity of asking the server for another one. For solving this problem, it could be useful to have the complete CC/PP (understanding this as an acronym not an specific format although the standard seems quite good) as metadata in the presentation file, so we could reason if that presentation really suits our CC/PP profile (we should have metadata that described the generator interface for automating the system that asks for another presentation).
Let's supose that a UFO (an alien spaceship for the purists) lands on Earth, and tries to ask our system a tailored multimedia presentation connecting with an alien device which uses an strange displaying (3D, inserting information directly to the brain (or whatever), ...). The system would always fail in its intention of creating a tailored multimedia presentation, as it has no idea of the existence of this kind of display. If the output had the metadata of the process associated, and the device could understand that metadata, it could generate a (in some ways) functional presentation with that. What we could get would be in a sense, complete independence from the display hardware. (Not of the software (or data) but as [M2] showed us, that is not a great problem).
The idea is that at the client-side we could follow the process of generation in the reverse order (bottom-up) to try to generate a functional presentation and as similar to the original as possible. So, we would try to understand the final form presentation. If we can't, then we will go up in the layers and we would get to the specific constraints. If we can understand this, we could generate a presentation displayable by us with the same positioning as in the original. If we cannot understand the specific constraints or they are not applicable to our device, we could go up in the layered system. We could get to the upper layer (RST in Cuyers) and we could generate a functional presentation from it but with our own conversion process, adapted to our device.
It's also important to see that perhaps the media items are not compatible with our device. With the metadata associated with those, we could generate a minimal "textual" (originated from text) representation of them, or try to look for an alternative but now compatible media item in the future (?) Semantic Web.
When a presentation is generated nowadays, all the information about the process of creating it is lost. In the final form we only get the final disposition of the elements. The idea is to make all this knowledge explicit, so that new presentations with the same meaning can be generated, still conservating most of the designers idea if possible (this leads to forward compatibility).
Another interesting question (at least for me) is the possibility of creating client-side the presentation from the rethorical structure. This could take us to the creation of homogeneous presentations from heterogeneous sources, and the elimination of links as an explicit element.
That's more or less what I've described. The concept itself is to mix the final presentation form with some more abstract data about the disposition of the elements (CUYPERS: qualitative constraints ), and the semantic relations between them (CUYPERS: rhetorical structure ).
[1] "Towards Second and Third Generation Web-Based Multimedia" Jacco van Ossenbruggen, Joost Geurts, Frank Cornelissen, Lloyd Rutledge, and Lynda Hardman.
[M1] Mars Attacks! Tim Burton, 1996
[M2] Independence Day Roland Emerich, 1996
It's clear that although the semantic relation between two media items can be the same, the type of the media items (video, audio, text, ...) determines the mapping from the
rethorical structure to the communicative devices. In Cuypers system we have a lot of hybridd media groups ( groups of media items of different type ) that require an hibryd
treatment.
The analisys of media types (pure and hybrid) and how they affect the Rhet-ComDev mapping could constitute a good project, with some practical work on Cuypers.
CC/PP is a W3C language created to provide an RDF-based framework for the management of device profile information ( to specify devices and user preferences ). A possible project could be an analisis of what it adds to the existing market of especifications, benefits, alternatives, ... Studying modifications needed in Cuypers in order to be CC/PP aware (or partially at least), and implementing them as a practical work and example of CC/PP deployment.