Use of RDF to specify accessibility status of pages. This continues what we were discussing last tuesday's joint meeting with AU.
Our action item there was to review the capabilities of RDF... at least at a high level... and then think of some concrete examples of what assertions we would want to write and what inferences we would want to extract.
Example assertions:
Example inferences:
Use of classes to define parts of a document (see discussion of "Grouping Links" at 31 July minutes.
We just have to skim to get general idea... you can pick and choose from...
There are a number of intros to RDF on that page. A few that caught our eye.. don't know if they are the best...
HB Decreases clutter, they're all in the HEAD element.
WL Could be like external style sheets. Clutter is just link to RDF document.
LK In CMN example, when he refers to a resource, he spells out the full URI each time. With namespaces you could define an abbreviation that refers to the URI.
If you use a particular set of names a lot you can use those as the default names. As far as the end user is concerned this is all transparent.
WL I've started a new web site called rdf.pair.com/intro.html I would like some help on quotes and references. "Sipping from a firehose."
LK A quote on the page says that the difference from previous technologies is that previously centralized and now decentralized.
WL A zillion accessibility standards, who do you trust? How do you check that you are not being spammed by a porn site? The naked lady site put every word in the dictionary in their meta info, the search engine doesn't know the difference. Need digital signatures.
LK Let's come up with a bunch of RDF statements that we would like to see put on the Web then decide what to do with them. The machinery for doing stuff with this is being experimented with. Several inference engines available none are Recommendation.
We can RDF statements that identify who said what. A machine searching out the info could determine who asserts it.
WL Before you go into doing that, look at the dublin core. They have 15 recognized things.
LK /* looking at the dublin core */ title, creator, subject, date, type, format, identifier, source, language, relation, coverage, etc.
HB Some of those can be multiple.
LK Many of these have obvious uses.
WL Rather than reinvent the wheel, let's see what they mean by that. As you get into it, then you can say, "this comes under that" We'll have to be immersed in this since the tools will use it.
LK looking at just "title, creator, and subject" how would we use these? I'm an evaluation tool, I don't look at the page, but what other authorities have said about a page, then I draw a conclusion about the page. Let's come up with examples of how these are used.
WL Has no alt-tags, alt contains the word ".gif".
LK How actually use.
WC Need for page level and element level.
LK A page where AFB has approved the alt-text, a deafness organization has approved the captions in the movie, etc. Some statements will apply to multiple elements.
WL We have lots of good ideas for how to use. But, XML is just 3 letters. There is time to digest this.
LK I agree, but I'm trying to get some concrete examples for how to use this stuff and what we're going to do about it. Are we going to define RDF predicates that people can use?
HB In the latest Dublin Core they have added qualifiers (refinements) to their elements. "for relation" "has part, is references by, format of, etc."
LK When I look at the Dublin Core elements, they instantly make sense when I want to apply to a book or article, I am having trouble seeing how they apply to what we're doing. I"m not saying they don't apply, but I don't comprehend it yet.
WL We have a lot of studying to do.
LK Maybe we haven't assimilated the information well enough, but I personally do not see how to apply.
WC Relationship.
HB Digital Talking Book using some, still in flux. Mostly SMIL.
LK In the expanded aspect, do these touch on internal aspects of a book? chapters?
HB If a book has a narrator, who narrator specific chapters, commentary.
WL There will be concordances for War and Peace just as with Bible. You better have a way to sift through it.
LK If an authority publishes a review of a web site or part of it, The Dublin Core provides us with the way to publish the commentary. Creator, Contributor. Commentator. You could have cross evaluation project to evaluate web pages. This describes the commentary itself.
WL This whole vein is that a lot of the meta stuff appears within a document, e.g., creator, date, etc. But it doesn't have to be. E.g. you can comment on the Bible without having written it.
LK Does the Dublin Core have special terms for works that are reviews of other works.
HB Don't believe so, no. "Is referenced by" and "references"
LK Are there various types of references?
WL You can create that on your own.
HB /* reads about references from dublin core */
LK Say there is a history book that uses HB's as a secondary source material. Or I could review HB's book and reference it. At this point in time, the dublin core does not distinguish between these references. I'm reviewing or making use of.
HB There are further text to use.
LK A web site is a review, we have used the dublin core on the site (contributors, date written, what referencing) it's referencing guidelines, the document it is reviewing, etc. Then the insides of the web page is making statements about the site. Like WL mentioned you can mention info about each alt-tag. RDF does have a way to assign numbers. We would have to define something new. But we could define something on a scale: poor, somewhat poor, good, very good, etc.
WL There was a public relations debacle that happened with PICS. PICS tends to be what we're looking for for evaluations.
LK Did they get into Dublin Core stuff?
HB It predated the Dublin Core.
LK For statements about individual elements, is it straightforward for every checkpoint to be obviously converted to an RDF statement.
WL WCAG checkpoints?
LK Looking at AERT.
WC Schematron. Would be interesting to look at it and fill in what's missing, e.g., is the alt-text accurate.
LK Looking at AERT, auditory descriptions...
WL When put it in, leave an audit trail that it is present.
LK Let's say, here is a page, it has a link to a real audio file. I want to say, "there is a text description for this audio file." I've prepared it, listened to the file, etc. I want to convert this to an RDF statement. "has-audio-description" "name of the file" "yes/no"
HB I can't believe the Daisy folks felt they needed 43 additions to the Dublin Core.
WL Somebody is saying the same thing you have said. They probably have not heard back yet.
LK What are those?
HB format generator, multimedia type, front page, page normal, etc etc.
LK Language is in the original set.
HB This might be a translation from the original.
LK Of that list, which most offends?
HB They all deal with document itself, no chunking.
LK We have consensus that there are useful things in the Dublin Core. Schematron is an initial cut at specific RDF statements. Question is: what additional statements do we need?
LK Here is an HTML document, here is an element.
HB It has class= already.
LK We can embed the comment in the doc or put it in external.
HB Or add an attribute value pair to existing element for which our commentary applies. CAST's Bobby puts something in new version of document as a placeholding mark (with hats).
LK Are the Bobby hats intended to markup the document for machine-readable tracking or just to communicate with the human.
HB Right now is informative to human. Used to link to more info.
LK Your talking about inserting new tag into new document. Say you have img with alt-text. Are you saying you would add attribute value to image tag itself?
HB That's one thought.
LK Another way?
HB XML comment convention. I would discourage it's use. Way to put in non-visual.
LK <!-- -->? Adding meaning to comments. Don't like that.
HB Basically trying to define an audit trail. It can either be in source document, where it can persist over revision or in concurrent document.
LK If in concurrent document, how does it make assertions about each element.
WL It's name contains something that that's what it is.
LK Here is an RDF doc, it says that a certain image lacks alt text. E.g. you can think of various ways to point to an image. XPointer "It's the 3rd child of 5th descendent" but what happens if someone asserts some material in the page. Then you break that relationship.
WL The revised document doesn't check out again.
WC Unless editor keeps track of things being inserted.
LK Fine if the editor is keeping track, what if external? There would have to be a diff between versions.
WC For image, could look at src.
LK There are heuristics. Sending diff to RDF to correct itself...
WL The way it was organized in the first place will dictate that. If originally done by machine, can be done again.
LK What kinds of standards would we have to define? Ideally want a diff tool that works on the DOM.
Action WL Start the basis of "What are the accessibility features of RDF"
Action WC: Check out schematron, how can we make it more precise?
WL I could try to arrange something in San Fran in November.
LK Piggyback off of ASSETS and CUU? DC in November.
Action LK post links to ASSETS and CUU
WC Who do we know in DC who could host?
WL Government organizations. NIDDR, US Mint, GSA, etc.
LK Just a room?
WC room, A/V, breakfast, lunch.
HB AOL is in the area.
Action WC ask JB who she knows in the area that might be able to host.
HB Also XML 2000 in DC in early December.
WC I am presenting there.
Action LK post link to XML 2000 to the list.
$Date: 2000/11/08 08:17:43 $ Wendy Chisholm