Agenda for March 27

Time of meeting is 9:00AM EST, 14:00 UTC, duration is 1 hour
Zakim Bridge +1.617.761.6200, conference 2742 ("ARIA"), IRC channel is
#svg-a11y

minutes from the March 20 meeting:
http://www.w3.org/2015/03/20-svg-a11y-minutes.html


Agenda

We will be looking at IBM RAVE sighted chart navigation.
web meeting link https://apps.na.collabserv.com/meetings/join?id=2897-5157
I will post the web share password in IRC.
You may be asked to download a plugin for the meeting. I will start the web
share 10 minutes before the start of the meeting to allow you time to
download and check that it works before the meeting.
I believe the web share works well with Firefox and Chrome. I am not sure
about other browsers.

I've included the main concepts used in RAVE chart navigation for sighted
users below. I hope to demonstrate the concepts using the chart listed with
Chart: . This should match the SVG, I sent to Chaals and may be posted on
the wiki. The concepts used in sighted user chart navigation are marked
with Concept:. An Interesting fact: is observation or implementation
detail.

RAVE sighted navigation for SVG a11y tf


Chart: scatter2


Concept: feature order


Feature order came from the National Center for Accessible Media.  Order -
title, axes, legend, data elements, reference items


Interesting fact: We always navigate to the Y axis before the X axis as the
Y is the dependent axis (even if you are doing 3-D or have clustered axes).


Interesting fact: We consider grid lines reference items - so they come
last.


Concept: Sighted users use arrow keys to move within a feature


Interesting fact: We let (non polar) axes roll over as a convenience to the
user.


Interesting fact: On data elements we don't roll over. Rolling over on data
elements would make it to hard for a user to get to the farthest item
without popping back on the other side.


Concept: Sighted users need to jump gaps. We use the arrow key + shift to
let them have a stronger move in the direction.


Chart: boxplot


Concept: Groups and items may have parts that have semantic meaning


A box plot (aka box and whisker plot) has 5 parts + (optional outliers).
The center line is the median. The box is usually defined by a percentile
value, but quartiles, standard deviations and sigma values may be used.
Likewise the fence (i-beam) is a broader percentile (std dev or sigma). If
there are sample members outside the fence they are outliers and may or may
not be shown. If shown they are shown as individual points.


Interesting fact: In RAVE we break a box plot into three parts. The fence,
box and median - since they are larger and larger ranges and the ranges are
symmetric.


We use the spacebar to cycle between parts. Shift spacebar cycles in
reverse order.  If an axis has a title, navigation treats the axis title
and the axis tick labels as parts to cycle through. The title getting focus
first, then hitting the spacebar would switch to the tick label parts.
Movement between axis ticks uses the arrow keys.


Chart: stackedColumn


Concept: Groups


When you have groups, you can think of the feature as a tree. You use arrow
keys to move within your branch level. Use the spacebar to move down a
branch level, spacebar + shift to move up a branch level.


Interesting fact: Our first attempt did not use tree style movement for
groups and gave bad behavior with stacked charts.


Chart: sierpinski5


Concept: Arrow key navigation (without the shift key) needs to follow the
users perception of neighborood. That is - navigation should not jump out
of your visual neighborhood - just because a point is farther/lesser in
X/Y.


Concept: Sighted users must be given more control than just 4 directions.
They need to be able to hop gaps and reach islands and jump out of
neighborhoods to reach other neighborhoods.


We provide arrow keys for moving within a chart feature. To provide a
stronger move which may bypass closer items but that are not so closely in
the arrow direction, we use the arrow key + shift. Both a normal strength
directional move, which keeps you in the neighborhood and a stronger move
which can get you to islands and allows you to jump gaps are needed.


Chart: clusterBar3levels


Concept: Multi level groups


Interesting fact:With clustering you can have multiple (nested) axes


Interesting fact: We don't use group behavior on axes. Axes can be put on
either side of the chart. So you could request an inner axes be put on the
opposite side of an outer axes.


Interesting fact: in AT supported navigation, groups are simply branches in
the tree. And we provide summary information (totals) at the branch level.


Chart: tree


Concept: movement using relationships


Interesting fact: When a user wants to navigate using relationships they
have to tell the widget, via a keystroke to go into "link" navigation mode,
they can still use arrow keys like normal too.


Interesting fact: On this chart there is no way to reach "2" without using
the arrow keys.


We highlight potential targets. Forward targets and backward targets are
highlighted differently. The target path is shown in thick orange to make
it stand out. The target has a wider black boarder. Products can set
different highlighting if they wish.


We provide keys to cycle through the potential targets. You can go forward
through the list or backward through the list. We provide a key to trigger
selection of the target.


Chart: 3graphs


Concept: there can be multiple graphs in the graphics, you need to be able
to move between them


Concept: Graphs should respect point of regard. So when you move between
graphs, you should go back to where you were before.


Interesting fact: we use the arrow keys to move between graphs, but we have
to tell the widget that we want to move between graphs. When we want to
move inside a graph again, we have to tell it we want to move inside the
graph


Interesting fact: In RAVE two graphs don't have to have anything in common,
they can just be in the same SVG.


Chart: chord


Concept: Sometimes you have to figure out what should be done


Interesting fact: Chord charts are like the Mobius strip of charts. There
is no beginning or end. Each category has a group of ribbons. A ribbon
usually belongs to two groups (categories), so you can't represent this
chart as a tree.


Chord charts show relationships between categories. The categories can be
the same type of thing (ie countries - look at chordLayout_singleCategory).
Or categories can be from different things (ie severity and violation) - as
long as  pair can be created.


In a chord chart, the categories are usually put on the x axis and wrapped
into a circle. The inner part of the circle is filled with ribbons linking
the two categories and defining the pair. (A ribbon may connect to the same
category twice instead of two different categories.) The ribbon width
usually indicates the magnitude of the pair link.


The outer circle of categories are navigated using the arrow keys.


The category is a group and we use the spacebar to jump to the ribbons
belonging to the group. This is the same behavior as clusters and stacks.
Arrow keys move between ribbons. You can jump to the other category parent
by using the same key for selecting a link/relationship target.  We just
figured out that would be a reasonable thing to do.




                                                              
                                                              
                    Regards,                     Fred         
                                                              
                   Fred Esch                                  
       Accessibility, Watson Innovations                      
    AARB Complex Visualization Working Group                  
                     Chair                                    
        W3C SVG Accessibility Task Force                      
                   IBM Watson