- From: Al Gilman <asgilman@iamdigex.net>
- Date: Thu, 10 Jul 2003 12:58:36 -0400
- To: Marja-Riitta Koivunen <marja@w3.org>
- Cc: public-wai-rd@w3.org
At 06:30 PM 2003-07-09, you wrote: >Hi Al! > >It was nice talking to you at Budabest. I remember that you had some good >ideas for visualization and also links to research material. > >We are preparing for a topic on visualization (from one angle) and I wrote >a first draft explaining it: > >http://www.w3.org/WAI/RD/2003/06/visualization/visplan.html > >Would you be so nice and send me or rdig list (public-wai-rd@w3.org) some >information about the links that you think would be suitable for this kind >of an event? > >Also if you have other ideas could you send them to the rdig list? [Caveat: I have not gone through the resources you cited. You may have a lot of this already. But here is a dump of some things I have been aware of that you may or may not have. -Al] You raised two aspects: reducing dependency on visualization to afford effective non-visual (or limited visual) access, and increasing the application of visualization methods to address cognitive and language barriers. This reply looks primarily at the first aspect: deconstructing the visualization to afford more universal access. But the skill developed in this pursuit will have lots of spillover into understanding how to make more and better use of visualization in more web content. Key points are: 1) Perform a "universal design" transform on the topic statement. FROM: "access to visualizations" -- document focus, assumes sequential derivation from data to visualization to accessible visualization TO: "access to information that benefits from visualization techniques" -- service focus, allows parallel derivation of the accessible dialog/view, which may or may not be integrated with the visualization. 'Visualizations' are just a view of some information that happens to work well for visual interaction. A deeper model of the information being served will facilitate the synthesis of an effective interaction for other delivery contexts, whether this is implemented as an enriched visualization or an alternate modality of dialog entirely. I have over and over written that we need to stop worrying about accessible maps and win by ensuring that accessible "location-related information services" are available for the same information as visualized on maps. For this workshop we need to re-orient the problem statement in the appropriate generalization of this remark. 2) Recognize that the presentation only has to sustain orientation to the conceptual space in which the information is distributed, not present the full dimensionality at all times. At least for scientific visualizations, the information to be accessed is a continuous distribution across some geometrical space. Verisimilitude in the interaction view is helpful in orienting the user to this conceptual space, but need not be complete. The relationship of the interaction spatialization to the conceptual space in which the information is distributed is that of a tangent space to a smooth manifold. The presentation only has to present changes in the orientation of the view well enough for the user to track an orientation in more dimensions than what the presentation covers. The success of the "Audio Doom" game with blind children illustrates this well. It is an extreme case. A more 'central' case is the Timesearcher application from Ben Schneiderman and Co. (see more below). 3) Meet the visual interface designer half way at the "scene graph" or "Last graph before pixels" interface. The W3C DOM of a Z39-86 digital talking book is an accessible information model, but it throws away too much of the application-domain geometry to be a reference model of choice here. Looking for a W3C DOM throws away the application-domain geometry that we need to communicate to the user. Think as a key target for an open interface in the architecture of computers and their cross-application software. The DOM is document-oriented; we need to have more attention paid to the scene graph that drives the screen or other multimedia or VR display. 4) Work the GIS application: it's low hanging fruit and a model that can be proliferated into other data browsing situations where there is significant geometry in the notional domain over which the information is spread. Why is this low-hanging fruit? Because the geometry here is of low dimensionality and is commodity knowledge among users. They already have the concepts; all we have to do is connect with what is already in the user's brain. And there is an immense amount of verbally articulable information, here. 5) Study the time-series browser from Ben Schneiderman's Lab. This is a good, working example of the kind of thinking that we need to teach people. Timesearcher http://www.cs.umd.edu/hcil/timesearcher The key fact about this application is that while the visual user feels that they are directly manipulating the display, looking at the processing flow they are actually tuning the input port on the view-extraction transform. So they maintain continuous orientation to the terms of the view effortlessly. Al PS: Further References: General background: view agility depends on thorough modeling: UD/DA considerations for The Grid http://trace.wisc.edu/docs/ud4grid/ Presentation/Interaction view as tangent to data space: "Audio Doom" study http://216.239.37.104/search?q=cache:rGgQ0yfVi5YJ:www.dcc.uchile.cl/~jsanchez/Pages/papers/virtualenviorment.pdf+3D+audio+doom+&hl=en&ie=UTF-8 A rich GIS website with lots of user control over the information view. Useful as a case to discuss in terms of adaptive views for different user situations. SanGIS, San Diego City and County Maps http://www.sangis.org/ Scene graph rather than document model as the central model class for studying universal design transformation of visualization applications: http://lists.w3.org/Archives/Public/w3c-wai-ua/2002AprJun/0102.html Goodies: It's possible to improve on the localization of audio presentation by employing more of the frequency spectrum: ineluctable sound from Leeds http://news.bbc.co.uk/hi/english/sci/tech/newsid_1454000/1454737.stm There are strong, open-source tools such as VisPACK supported by high-value applications: BISTI Program of Excellence for Computational Bioimaging and Visualization http://www.sci.utah.edu/bisti/static/ >Thank you! > >Marja >
Received on Thursday, 10 July 2003 13:42:58 UTC