W3C home > Mailing lists > Public > w3c-wai-gl@w3.org > January to March 2002

Fw: Checkpoint 3.3

From: Lisa Seeman <seeman@netvision.net.il>
Date: Wed, 13 Mar 2002 09:41:22 -0800
To: "_W3C-WAI Web Content Access. Guidelines List" <w3c-wai-gl@w3.org>
Message-id: <016901c1cab6$4b318640$2991003e@dev1>
Forwarding this to the list. very cool
----- Original Message -----
From: "Jo Miller" <jm@bendingline.com>
To: "Lisa Seeman" <seeman@netvision.net.il>
Cc: <wendy@w3.org>; <jasonw@ariel.ucs.unimelb.edu.au>; <GV@TRACE.WISC.EDU>
Sent: Tuesday, March 12, 2002 3:39 PM
Subject: Re: Checkpoint 3.3


> Hi Lisa,
>
> Thanks for your comments!
>
> Do you have any objection to moving this discussion to the mailing
> list? If not, I would like to take it there so that the others can
> participate if they choose. (Maybe you copied the list already and
> the message just hasn't come through yet. I've noticed quite a delay
> lately on listserv message distribution.)
>
> >All in all,  I enjoyed reading this one, (really pointed out what I had
> >missed.)
> >comments in line
> >
> >>  Broadly speaking, written content can be divided into two types:
> >
> >for text in images we had said "unless the main purposes of the text is
non
> >semantic" or something like that. That covers both exceptions  - if it is
a
> >citation or if it is truly art. It also allows a web author to create an
> >exception through redundancy
>
>
> I'm sorry, I don't think I quite understand what you are saying here.
> Could you clarify for me please?
>
> If you are describing art as a sub-category of "content that cannot
> be edited," then I agree, pretty much. Both would be exempt from some
> of the success criteria that apply to "editable" content. However, I
> can see some reasons why we might want to mention "art" explicitly as
> a sub-category, to avoid potential confusion:
>
> 1. Some art (poetry, let's say, or short stories) may be created
> solely for the web and may not be published anywhere else. If the
> poet/author is publishing her own material online, she might wonder
> whether or not her work qualifies as "uneditable" (after all, the
> poet could, in theory, rewrite her own poems). Charles's proposed
> category for "artistic" content would eliminate this potential
> confusion.
>
> 2. Requiring web authors to provide an apparatus to help all users
> understand artistic content would be, in many cases, unreasonable,
> and might deter people from publishing artistic material at all. I
> believe Charles or Graham raised this point in an earlier discussion.
> If I were required to explain all the unfamiliar terminology in a set
> of Shakespeare sonnets, or provide an interpretive framework for
> works of James Joyce or Gertrude Stein, I might simply elect not to
> post that material. So we may want to consider how to limit the
> burden that we place on people who want to publish artistic content
> on the web.
>
> We want to make it clear that "uneditable" content is exempt from
> some requirements of 3.3, but not from all of them. There are still
> things that a web author can do to help readers understand uneditable
> content. He should provide accurate, unique page titles, for
> instance, and summaries. He can also provide appropriate structural
> markup, glossaries of jargon, explanations of figurative language,
> etc.
>
> The question is, how much supplemental material should a web author
> be expected to provide for artistic content? (For instance, is it
> even possible to nail down the meaning of figurative language in a
> poem, which may be subtle, multi-layered, and open to several
> different interpretations?)
>
> The "art" suggestion was Charles Munat's, though, and I'm sure he can
> explain the reasoning behind it better than I.
>
> >  > Questions: What is being summarized/outlined? A page? A set of pages?
> >>  A section of a page? Do we simply rely on the web author's common
> >>  sense and trust him not to over- or under-summarize?
> >
> >the upper limit could be every page can be summarized. For very simple
> >pages,
> >this can be done through a good title. (e.g. something.com's products)
> >I can sometimes be at a relatively simple page, and it is something of an
> >intellectual exercise to know what this page is actually about.
> >
> >Many sites have the same title for all pages. this is often less then a
> summary of the page, but a marketing thing for search engines.
>
> Do you have any suggestions for capturing these scenarios in the
> success criteria? If a title can sometimes suffice as a summary (and
> I think this is a good point -- sometimes an accurate title is
> plenty), is there a way to say so without giving people the idea that
> a good title is always an adequate substitute for a summary or
> outline?
>
> Perhaps we should break page titles out as a separate success
> criterion. After all, good page titles ought to be a requirement for
> all pages. Then we could further refine the success criterion that
> requires summaries or outlines. For instance, we could say that pages
> containing more than [x amount] of text should be accompanied by a
> summary or outline, but for shorter pages it is not necessary. I
> don't know what the limit would be. Five paragraphs? One screen-full
> at 800 x 600 in default font size?
>
> >
> >>  "Use short paragraphs with no more than one main idea per paragraph."
> >>"short" is a relative term (how short is short?).  We are leaving
> >>  it to the author to decide what is appropriate for her content.
> >
> >We can say that for most content five sentences is the maximum.
> >If your content is exceptional, such as...(I have no idea when this is
> >truly impossible) were short paragraphs are not possible, then use as
short
> >as possible.
>
> If so, then this criterion would belong in "advice."
>
> >
> >
> >Can you give me an example of a paragraphed of more then five sentences,
> >that
> >could not be broken up?
>
>
> I can. This paragraph is an example. It is short. In fact, it is only
> two lines long. It makes only one point. Yet it does do in six
> sentences.
>
> Remember, our advice to break up long sentences into shorter ones
> will lead to more sentences per paragraph. Breaking such paragraphs
> up at arbitrary points could simply create a staccato effect without
> improving comprehensibility.
>
> Don't get me wrong -- I agree that "try to keep paragraphs to fewer
> than five sentences" is excellent advice, especially for web writing.
> I agree that it would improve the comprehensibility of the majority
> of web pages. I just doubt that we can give it normative force
> without potentially doing more harm than good. I would leave it in as
> advice.
>
> Here is an additional consideration: I work with web authors and
> content providers on a daily basis, and unfortunately I can foresee
> an unwanted outcome of the "five sentences" criterion. Many authors
> -- especially those who care only about technical compliance with
> regulations -- will simply count off sentences, 1-2-3-4-5, and insert
> an arbitrary paragraph break at that point. (Gian also works with
> these folks and I think she might agree with my observations here.)
>
> This bean-counting approach will not result in coherent paragraphs
> with one idea per paragraph. In fact, it will probably produce
> paragraphs that are longer than they need to be. In most cases, FEWER
> than five sentences per paragraph is ideal. Yet a sentence-counter
> would simply see that his paragraph came in under the magic "5"
> number, and would say, "Good enough! I pass!" He would not ask
> himself, "Is this paragraph as concise as it could be? Does it make a
> single point in a straightforward way? Is it coherent and logical?
> Could it be shorter?"
>
> Therefore, I would far rather keep the focus on "short paragraphs,
> one idea per paragraph" than turn authors into automatic
> sentence-counters. I think this approach, though admittedly less
> concrete, would produce better results. Still, I see nothing wrong
> with telling people (informatively) that five sentences or fewer per
> paragraph is usually a good length to aim for.
>
> >  >
> >>  "Highlight your document's structure and its key points with
> >>  appropriate markup (e.g., headings and subheadings, emphases, lists)
> >>  to facilitate skimming and reading."
> >>
> >>  Testable? I'd like to think that this criterion would pass the "eight
> >>  reasonable people" standard, but what does the rest of the group
> >>  think? There's certainly a subjective element here, and the question
> >>  of sufficiency arises. When has the author done enough to claim
> >>  conformance?.
> >>
> >
> >The test I think we had marked out for it was if you just have the
highlight
> >content. Does the document flow? are any key points not represented?
> >
> >That is human testable.
>
> It is particularly testable if the document has gone through an
> outlining stage. Then one could ask whether all main points in the
> outline are emphasized through structural markup in the final written
> document.
>
> >
> >>  People who
> >>  understand the "why" are far more likely to implement this advice
> >  > successfully.
> >absolutely the why is very important. we will get there.
> >
> >>
> >>  "Provide definitions for any jargon or specialized terminology used
> >>  in your document."
> >>
> >>  Questions: What are the acceptable ways of fulfilling this
> >>  requirement? Would a link to a glossary of specialized terms suffice?
> >>>  "Provide explanations of figurative, metaphorical, or idiomatic uses
> >>  of language (for example, 'haven't seen you in donkeys' years' or
> >>  'the sight tore my heart out')."
> >>
> >>  Useful? If it is true that there are cognitive disabilities that
> >>  prevent people from understanding figurative language, then yes.
> >
> >There are. I wrote a few email to the list on this.
> >
> >>  Testable? Drawing the line would be a challenge here, and deciding
> >>  what needs glossing and what does not might require an expert.
> >>  Implementation is also a big enough pain in the ass that, practically
> >>  speaking, authors are likely to ignore this criterion. That decision
> >>  is not our problem, as long as it does not lead them to ignore 3.3
> >>  entirely. "Idiom" would cover slang, as well, by the way.
> >
> >we had using a literal translation tool to a foreign language, and then
> >doing
> >a literal translation tool to go back to the original language. If the
> >meaning is different then that is how some people will read
> >it.
>
> I do not believe these tools are equal to that task yet. I have spent
> many a giggly afternoon on the phone with friends, feeding sentences
> into Alta Vista's Babel Fish and translating them back and forth
> between English and German, English and Japanese, etc. The results
> are invariably hilarious, whether the original sentence contained
> idioms and metaphors or not.
>
> In several more years, translation tools may be able to handle the
> syntactic differences between languages more smoothly. At present,
> though, their tendency to translate word-by-word will produce an
> overwhelming surplus of false positives if we try to use them for
> this kind of testing.
>
> Not that it wouldn't be fun.
>
>
> >  >  Substitutions for mine are eagerly solicited.
> >go fly a kite (not you sir)
>
>
> <grin>
> Good one.
>
> >  >
> >>  OK, these last two bring us to the success criteria that address word
> >>  choice--a thorny thicket if ever there was one....
> >
>
>
> >I think that the more commonly used a word is, more easily it is
understood.
>
>
>
> Not necessarily. Take, for example, the scariest sentence commonly
> heard in the English language:
>
> "The plane will be in the air momentarily."
>
> Or the most disgusting:
>
> "Take this diet supplement and watch the pounds LITERALLY fall off."
>
> Ewww. Seriously, though, we cannot reliably predict which words users
> will understand. Vocabulary is too quirky and unpredictable a thing
> and, in my experience, seems to depend largely on factors other than
> cognitive ability.
>
> Furthermore, we would probably be making too many implicit
> assumptions about a site's audience, and we would certainly be
> impinging on the author's right to choose what is "appropriate for
> the content." Hence I think suggestions about word choice belong
> under "advice." I see no useful way of making them normative.
>
>
> >
> >To make this easily testable we need a Thesaurus / dictionary that ranks
> >word
> >usage.
>
>
> In every language? And is a web author expected to consult this
> thesaurus for every word?
>
> Believe me, I would be thrilled beyond words if authors paused
> between every sentence to consult a dictionary and ponder their word
> choices. That would be my wet dream (a figurative use of language
> that I would need to gloss, were I claiming conformance!). I would
> also love it if we had peace in the middle east and fuel-efficient
> SUVs. All of these things are equally likely to come to pass in my
> lifetime.
>
> Keeping a dictionary on the desk and consulting it routinely seems to
> be one of the habits people find most difficult to adopt, for some
> reason. Perhaps it's something to do with misguided pride. Perhaps
> it's just laziness.
>
> But even if web authors could be persuaded to adopt this habit, I
> would still oppose, for reasons stated already, any success criterion
> that limited the lexicon on which authors were permitted to draw.
>
> If we want to include use of such a dictionary (assuming one exists)
> as a piece of advice, I'd say that's probably fine -- though I do
> think that choosing the most accurate word available nearly always
> produces better communication than choosing the commonest word
> available.
>
> Let me clarify, at the risk of repeating myself: even if we could
> come up with a testable way of drawing the line -- e.g., "Thou shalt
> not use a word with a so-called difficulty rating of less than three"
> -- I would not support such a restrictive and probably
> counterproductive success criterion. Advice, yes, criterion, no.
> Authors need to be free to choose the RIGHT word to express their
> ideas. We cannot demand that they avoid the best word and substitute
> a less accurate word, simply because the right word may be one that
> is seldom used. Doing so would impoverish communication and do a
> disservice to writer and reader alike. (Besides, the tremors caused
> by Flaubert violently spinning in his grave might cause major natural
> disasters in Normandy.)
>
> It seems to me that asking authors to provide a link to a dictionary
> (either internal or external) is perhaps the best practical way to
> ensure that readers have the resources they need to understand a
> text, regardless of which words they already understand. One reason I
> like this solution is that it offers greater choice and flexibility
> to the users.
>
> By the way, the National Cancer Institute has an interesting
> implementation of a dictionary for specialized terms. [1] Obviously
> there are a lot of these special terms in Cancer Land. The reader can
> choose to have an NCI web page displayed in "normal" mode or
> "dictionary" mode. In dictionary mode, specialized terms are
> highlighted and hyperlinked to definitions from a variety of sources.
> I believe the system uses pop-up windows; it's been a long time since
> I played with it.
>
> I'm sure this approach is not unique to NCI; I mention it simply
> because it seems like a good solution for sites that have to use a
> lot of jargon or unfamiliar terminology. Users who do not need the
> dictionary can turn off the highlighting so that it doesn't distract
> them, and users who do need the definitions have very easy access to
> them. Each user gets to choose which presentation he prefers.
>
> Best,
> jo
>
> [1] See http://plan.cancer.gov/scipri/genes.htm and click on "Define
> terms on this page."
>
> >
> >I had understood that we could consider something testable even if the
tool
> >was not yet created for it, but could be created and assume that Bobby or
> >others will follow suit, even if there is some time laps.
> >
> >
> >All the best
> >Lisa
>
Received on Wednesday, 13 March 2002 02:49:55 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:47:18 GMT