Re: EvoGraphs plug-in -- a relevant practical implementation

The axis titles and overall title are included as chart options (the code
snippets are hidden in the details + view farther down the project web
page).  There are also options for which summary statistics to include.

It's certainly not generalizable to more complex charts at this point: it
only works with a single set of categories and numbers.  I was more
interested in the choices relating to presenting the data as text.  Ather
mentioned in particular that users had found the story-telling text
description easier to understand than a visually hidden row/column table,
which is what Google Charts uses.

ABR

On 25 September 2015 at 08:18, Fred Esch <fesch@us.ibm.com> wrote:

> Amelia,
>
> The goal is nice, but what is shown on the web page appears to be an
> experiment and not something we can learn from.
>
> If I read it correctly, the input data is a single JSON array. Where every
> other item is a category or value. There is no name for either data
> 'column', no declaration of data type and no encapsulation of rows.
>       data:   [                       //"category", "value" format
>                 "Cars", 12,             //don't forget the 'commas'!
>                 "Cars", 23,             //duplicate values are merged
>                 "Trucks", 45,
>                 "Vans", 22,
>                 "SUVs", 30,
>                 "Jeeps", 5
>                 ],                      //don't forget to close the ']'
>
> Also looking at their examples on the page, nowhere does not state that
> the categories are 'vehicles' and it does not indicate that the number is
> 'repairs'. So it is difficult to see how they could get their description.
>
> This is a horizontal bar chart. Its title is Horizontal Bar Chart. It
> contains 5 *vehicles*. Trucks is the first *vehicle* and has the largest
> total number of *repairs* of 45, which is 32.9% of the total data set.
> The next *vehicle* is Cars with a total number of *repairs* of 35, which
> is 25.6% of the total data set. The next *vehicle* is SUVs with a total
> number of repairs of 30, which is 21.9% of the total data set. The next
> *vehicle* is Vans with a total number of *repairs* of 22, which is 16.1%
> of the total data set. Jeeps is the last *vehicle* and has the smallest
> total number of *repairs* of 5, which is 3.7% of the total data set. The
> mean is 27.4. The standard deviation is 13.5.
>
> This description is not explorable, that is there is no way to find out
> about Jeeps without hearing about the other vehicles first.
>
> In each example chart ids are reused so the id 'inner-graph-area' (and
> several others) appears several places on the web page, which is a
> violation of WCAG 2.0 (4.1.1 Parsing). I am not impressed by how much stuff
> on the page is stuffed inside elements with aira-hidden="true" and don't
> think these folks show any expertise in creating an accessible chart
> rendering engine.
> Regards,
>
> Fred Esch
> Accessibility Focal, Watson Solutions
> AARB Complex Visualization Working Group Chair
> W3C SVG Accessibility Task Force
> [image: IBM Watson]
> [image: Fred]
>
> [image: Inactive hide details for Amelia Bellamy-Royds ---09/25/2015
> 12:19:55 AM---Hello all, There was a very interesting presentation]Amelia
> Bellamy-Royds ---09/25/2015 12:19:55 AM---Hello all, There was a very
> interesting presentation at the Graphical Web conference
>
> From: Amelia Bellamy-Royds <amelia.bellamy.royds@gmail.com>
> To: SVG-A11y TF <public-svg-a11y@w3.org>
> Date: 09/25/2015 12:19 AM
> Subject: EvoGraphs plug-in -- a relevant practical implementation
> ------------------------------
>
>
>
> Hello all,
>
> There was a very interesting presentation at the Graphical Web conference
> today from Ather Sharif at evoXLabs, Saint Joseph's University.
>
> He has been working on scripting tools to create screen-reader friendly
> basic data visualizations.  The product, evoGraphs, currently has two main
> features:
>
>    - It generates basic SVG bar charts or pie charts, using an interface
>       similar to the Google Charts API.  However, the SVG that results is
>       structured much better than Google Charts' output for alternative text.
>       The alt text includes basic full-sentence descriptions and annotations of
>       the data highlighting features such as max/min values, or overall summary
>       statistics.
>       - It can process an existing SVG chart, identifying grouped labels
>       and title/desc elements, and reformat it to the screen-reader-friendly
>       structure based on as much data as it can infer from the existing markup.
>
> In other words, it overlaps quite a bit on the work we have been doing or
> discussing.  I encourage you all to take a look at the choices they have
> made.  It currently only applies to very simple visualization structures,
> but it covers many of the issues about to what degree can software
> automatically convert data into textual descriptions.  Importantly,
> evoXLabs has actually been doing some user testing on their projects, which
> is something we haven't had much of a chance to do with the task force yet!
>
> The main project link:  *http://www.evoxlabs.org/whitecane/evographs*
> <http://www.evoxlabs.org/whitecane/evographs>
>
> And yes, Doug did hint that perhaps Ather may want to get involved in the
> Task Force in some way, so maybe we can hope for more feedback on *our*
> work based on his team's experience.
>
> Best,
>
> Amelia
>
>