- From: Will Pearson <will-pearson@tiscali.co.uk>
- Date: Wed, 15 Dec 2004 15:14:57 -0000
- To: <wai-xtech@w3.org>
----- Original Message ----- From: "Will Pearson" <will-pearson@tiscali.co.uk> To: "Al Gilman" <Alfred.S.Gilman@IEEE.org> Sent: Wednesday, December 15, 2004 2:49 PM Subject: Re: Accessing Graphics > > ----- Original Message ----- > From: "Al Gilman" <Alfred.S.Gilman@IEEE.org> > To: <wai-xtech@w3.org> > Sent: Wednesday, December 08, 2004 3:59 PM > Subject: Re: Accessing Graphics > > > > > > At 9:17 PM +0000 12/6/04, Will Pearson wrote: > > >Hi; > > > > > >Diagrams are a prevalent form of communication in contemporary > > >society. They are used to explain task sequences, convey concepts, > > >even design interactions between classes in a UML sequence diagram, > > >yet they remain one of the last frontiers in the world of accessible > > >information. This need not necessarily be the case, as after all, > > >diagrams are just a transport mechanism for meaning as much as the > > >words on this page are. > > > > But diagrams are frequently used when the plot of the story to be > > told does not follow a single, linear > > thread or loop-free topic tree. > > I agree. My comparison was just to illustrate the fact they both convey > semantic meaning, but use different forms of encoding :-) > > > > So accessing a linear narrative, or a linear-plus-grouping > > book-structured treatise does not give us all > > the precedents we need to deal with access to diagrams. > > > > We are familiar with a variety of navigation modes from > > book-structured documents: > > > > - serially through the whole thing in full detail > > > > - serially through action opportunities via tabbing > > > > - with seven league boots through titles of sub-topics of the current > > topic as in the DAISY table of navigation > > > > - table navigation, up/down left/right inside a regular grid of > > repetitive cells > > > > I think that in accessing diagrams we need to recognize that there > > are arcs linking the objects in the scene. Some of the objects are > > diagrammatically presented as connected or related, while other pairs > > are not. > > > > So there is a new sub-function involved in what I call "graph > > navigation" which is that there need to be facile means to discover > > and navigate to the strongly related objects, based on the currently > > focussed object. > > Yes. Poehlman and myself have been discussing this recently. Our latest > thoughts went into the keyboard exploration scheme in the SVG 1.2 feedback. > However, I'm not convinced this is the perfect method, as you're still > trying to squeeze what is a parallel, spatially based, system of encoding > through a linear, sequential channel. > > > > > > >There's three main ways in which I consider diagrams can be made > > >accessible. Each involves extracting the meaning from it's > > >diagramatical encoding, but each differ in where that decoding takes > > >place. > > > > > >A bit of communications theory. > > >There's various communications models used to explain how people > > >communicate with one another and technical communications systems. > > >One of these, which is classed as a transmission model of > > >communication, is Claude Shannon's 1948 model that featured in the > > >Bell Systems Journal of that year. Shannon proposed that there were > > >five stages to communication: > > >1. A sender considers the meaning to be sent > > >2. That meaning is encoded into some physical form > > >3. The physical representation of the meaning is transmitted to a > > >receiver, using physical communication channels > > >4. The physical representation is decoded to expose the transmitted > meaning > > >5. The receiver then absorbs the transferred meaning > > > > This is fundamental. > > > > We have this cycle now more integrated in the WAI public messaging. > > At least a toehold. See > > > > http://www.w3.org/WAI/intro/components > > > > But we still have to follow through on this principle. > > > > In WCAG, for example, it has to be clear that the > > author-through-server are responsible for delivering something that > > enables the user's control of presentation through the application of > > user-configurable transforms in the User Agent. > > > > >If we apply this to diagrams, the lines, colors, spatial > > >relationships are purely encoding, and are distinct and separate > > >from the meaning they convey. > > > > That's where I fall off the track. The use of 'purely.' Even in the > > business to business world of electronic document interchange, > > there's no purity to the encoding. And the encoding there is more > > consistent, more strongly controlled, than in the bulk of the > > X-to-consumer Web. > > > > The encoding is the relation or mapping between percepts and > > concepts. The percepts are not pure anything. One perceives what one > > conceptually expects. It's all part of a coupled, recursive process > > whether on the speaker's side emitting the 'communication act' or on > > the hearer's side building an estimate of what the speaker was > > thinking. > > > > >Therefore, to get at the meaning, all that needs to be done is to > > >decode the physical representation of that meaning. > > > > Once again, this exaggerates the precision of encoding in human > > communication. Natural communication is full of what we call > > allusion. You could call it fuzzy encoding. > > > > While on the one hand, I have been saying that we need to make the > > cycle Will described above the backbone of our model of Web > > communication, and present accessibility in that context, we still > > need to go a step further to recognize that web content is > > semi-formal. By this I mean that there are strict models for some > > aspects of what is being conveyed, but not alll aspects. > > > > A key plot point is that in Web communication there is a transform > > being done at the client side from the wire format to the physical > > presentation (and event acceptance) form. It is in many cases easier > > for the user to control this transformation being done at the client > > side than to reach our and perturb what is being done on the server. > > [But not always] > > > > >It's this decoding that causes problems in accessibility. > > >Psychology examines the process of receiving meaning in a bit more > > >detail. According to psychology, we first receive sensory stimuli, > > >which can be in the form of waves, particles or contact with other > > >physical objects. We then automatically group these into perceptual > > >groups, which in the world of diagrams would be the lines, shapes, > > >colors, words, etc. that form a diagram. The final stage in this > > >process is for us to cognitively associate meaning with those > > >perceptual groups. > > > > When the application is mailing the baby's picture to a grandparent, > > most of the message is in the image; the concept that this is their > > grandchild is part, but the smaller part. > > I agree, but how do they determine it's their grandchild? They would take > all the pixels that they receive as sensory stimuli, the brain would then > group them into perceptual groupings based on something like the Gestalt > laws, they would then recognise it's a baby. They would then > > perform this for all the components of the image, realising the person > holding the baby is their son or daughter, and then make the assumption that > it's their grandchild. I say assumption, as unless they've been told > otherwise there could be a number of reasons why their son or daughter is > holding a baby. > > So, the meaning is conveyed through the spatial relationships between > pixels, and at a more coarse granularity, the relationships between objects > in the image. > > > But in diagrams, the message is symbolic and we have lots of ways to > > represent or interactively browse said message. > > Yes and no. I think the same component operations are used for both > diagrams and images, but we perform the operations sub-conciously with > images. We associate some meaning with a group of pixels within a given > context, and we further enhance that meaning through other semantic encoding > channels such as color and spatial relationships. For example, the baby is > being held by a person, and we know it's a picture illustrating someon'e's > new born child. Regardless of how many other babies are in the image, we > know it's their baby as a result of the spatial relationship between the > baby and the person. This is true for diagrams, where you can gain > additional meaning from the spatial relationships between diagram elements. > > > > >Examining the psychological process of receiving information, > > >there's two main problematic areas for accessibility. Either people > > >can't receive the sensory stimuli due to physical, environmental or > > >other constraints, or they cannot cognitively associate meaning with > > >the perceptual groupings, which may be due to one of a number of > > >factors. > > > > FWIW my recounting of this tale is at > > > > http://trace.wisc > .edu/docs/ud4grid/#_Toc495220368 > > > > >Semantics can resolve both of these issues. If it's embedded as > > >part of the physical transportation medium, then the transferred > > >meaning can be reassembled in any form suitable for the user. This > > >could be a form that bypasses problems, be they physical, > > >environmental or whatever in nature, that prevent the user from > > >receiving sensory stimuli, or it could be a form adapted to allow > > >the user to cognitively associate meaning with the perceptual > > >groups, where they may have been unable to with the original > > >perceptual groupings. Most people are familiar with the fact that > > >some people cannot receive certain types of sensory stimuli, the > > >blind cannot receive light waves, the deaf sound waves, and so on, > > >or it may be in appropriate for people to receive certain types of > > >stimuli, well, you need to look where you're going when walking, you > > >may fall down some steps. However, accessibility goes further than > > >just dealing with issues of disability, be it permenant or physical, > > >the ultimate aim of accessibility is to ensure everyone can access > > >meaning. This includes adapting the encoding of the physical > > >representation, but not the type of stimuli used to encode it. For > > >example, a blue line would yield no meaning to someone unfamiliar > > >with the UK's Ordenance Survey 1:50000 maps, yet it represents a > > >motorway. This is because they haven't learnt the particular set of > > >symbolic encodings used in an OS 1:50000 sheet. Through the use of > > >semantic content adaptation barriers such as this lack of knowledge > > >of various symbolic encoding sets can be overcome. > > > > > >Semantic content need not necessarily be encoded in the physically > > >transported content, it can be gained after transportation. The > > >final stage of the psychology sequence involves associating > > >cognitive meaning with sensory stimuli, or in other words, > > >extracting the semantics from the content. This process can be > > >automated by intelligent agent software that have been taught the > > >encoding techniques used in a particular diagramatical context, and > > >this set of extracted semantics can then be reencoded as if the > > >semantics were originally embedded within the transported physical > > >representation. > > > > > >Finally, and to me the most fun, as I've been working on this in > > >industry, is adaptation of the sensory stimuli itself. This is only > > >suitable for those unable to receive the sensory stimuli for > > >whatever reason, and involves converting diagrams and images into a > > >form of sensory stimuli that the intended receiver can receive. > > > > Try ... converting a web dialog containing diagrams into an alternate > > dialog that communicates... > > > > A lot of what I have had to say about access to problematic > > presentations, both transit timetables and tax-preparation > > flowcharts, has focused on what is known in the trade as 'equivalent > > facilitation.' Using > > interaction as a resource to eliminate the need for the presentation > > of an acyclic graph as essential > > to the task at hand. Specifically, getting the user to input where > > they want to go, and from where, > > means that the server can present a short list of route plans each of > > which is a linear story, rather than > > a route map or a timetable that takes a lot of skill to navigate. > > Also the flowchart was an inferior way > > to explain the logical flow through the preparation of a tax form; > > whereas activating the individual decision > > questions with hyperlinks provided a superior explanation. The > > availablility of active navigation eliminated > > the need for flowlines in the graphic. The story unfolds through a > > dialog rather than by reconstructing > > it tracing a path through the diagram. > > I agree that this is a good way to display some content. However, it only > works when there's a clear sequence through the elements, and doesn't take > into account jumping from one point to a random point, which the user may > wish to do for various reasons, maybe context, maybe they've forgotten > something. It also fails to take into account document exploration. A flow > chart will show all actions at the same stage in the process, therefore the > user can "look ahead" and determine what the consequences of taking a > particular fork in the sequence would be. Therefore, it aids in the > decision making process, and this sort of decision making ability is not > found in conventional wizard style dialogs. > > > > http://www.w3.org/WAI/RD/2004/06/28-agenda.html > > > > But this thread should be about access to the diagrams, with a brief > > nod to the alternative dialogs for the > > cases where the authoring side should go there first. > > > > We should review the Bulatov work, where it gets us and what needs to > > be done next. > > > > http://www.svgopen.org/2004/papers/SVGOpen2004MakingGraphicsAccessible/ > > There's also Rotard, M and Ertl, T, Tactile Access to Scalable Vector > Graphics for People with Visual Impairment, In proceedings of SVG 2004, > (abstract) > http://www.svgopen.org/2004/paperAbstracts/TactileAccessToSVG.html > > > >There's also Grigori Evreinov's work on Spotti mapping. Spotti, whilst > intended to be a synthetic vision system, could be used for diagramatical > access without some of the anatomical limitations of tactation. Spott paper > at: > http://www.acoustics.hut.fi/icad2001/proceedings/papers/evreinov.pdf > > Will >
Received on Wednesday, 15 December 2004 15:12:47 UTC