Reading order SC (promote to L1 from L3?)

Becky's summary of remaining work on GL 1.3 noted an open issue
advocating promotion of our SC about reading order from L3 to L1, and
suggested doing some research. I took this as an action item.

SC 1.3.5 reads as follows:
<current>
1.3.5 When content is arranged in a sequence that affects its meaning,
the sequence can be programmatically determined.
</current>

Here's my report. We'll discuss this on our Wednesday call and make a
recommendation to the WG.

<report>
I did four searches using Google Scholar:
1. "reading order" and dyslexia (13 hits)
"reading order" and blind (99 hits)
determine and "reading order" and OCR (153 hits)
determine and "reading order" and "Web page" and automate (21 hits)

My aim was to find information that would help us close the issue of
whether to promote our SC about reading order (currently SC 1.3.5) to L1
or L2 or leave it at L3. Also, if we do decide to promote it, should it
go to L1 or L2?

Conclusion: Promote this SC to L1.
Rationale
(1)	Reading order problems can create serious accessibility
barriers. 
(2)	It is very difficult for automated systems to determine reading
order from complex layouts; contemporary AT does not do it
(3)	Authors can easily control reading order without constraining
the default visual presentation (at least for technologies that work
with style sheets)

Discussion
The literature confirms that determining appropriate reading order is a
difficult challenge for both human readers and automated tools.
"Fluency" of reading is lost when readers have to focus on the
mechanical dimensions of the process, for example if they have to spend
energy decoding individual words or perceiving the reading order.  This
loss of fluency can have a major impact on ccomprehension.

Advanced document scanning systems for digitizing print documents employ
document scanning, image analysis, and optical character recognition.
The system divides the page image into zones, determines the order in
which the zones should be presented, and performs OCR to convert the
image into digital text; see "Document image analysis and understanding
R&D" (2001),
http://archive.nlm.nih.gov/pubs/reports/mars2001-DIAU/mars2001-DIAU.php#
a4 for a summary of approaches to page segmentation (zoning). All three
processes may require significant manual intervention. These
interventions help to improve performance as the system "learns"
layouts, fonts, etc
Screen readers do not analyze layout. They simply present content in
source code order and do not give users advance indication about the
content or how it is organized. SC 1.3.5 addresses accessibility
problems that occur when the user agent is not able to recognize which
parts of the content require continuity to be intelligible and which
parts can or should be presented separately.  A paragraph that flows
from the end of one column to the beginning of the nextis an example of
content that requires continuity, while the text at the beginning of the
first column must not be treated as continuous with the text at the
beginning of the second column. Other examples of content that should be
presented separately include  main content and navigation bars; a
sequence of paragraphs within an article and sidebars that present
related content which does not belong in the same sequence as the
paragraphs. In other words, if the user agent renders adjacent content
as continuous content, users will be unable to perceive a reading order
that makes sense.

In these cases, authors must specify a reading order and expose it to
the user agent. 

Some authoring tools, including Acrobat Professional 7.0, include
features specifically designed for this purpose. Some evaluation tools,
such as WAVE, report the linear reading order of HTML content.

</report>

"Good design is accessible design."

Dr. John M. Slatin, Director 
Accessibility Institute
University of Texas at Austin 
FAC 248C 
1 University Station G9600 
Austin, TX 78712 
ph 512-495-4288, fax 512-495-4524 
email jslatin@mail.utexas.edu 
Web http://www.utexas.edu/research/accessibility

Received on Friday, 30 December 2005 20:09:39 UTC