[BlindMath] Is it possible for screen reader users to do data science?

Tue Jun 27 15:53:19 UTC 2017

Hi All,

Thanks for all the useful information - I'll take a look at the tools
mentioned.

@Jonathan: I'll not be able to attend unfortunately (NFB's at the other end
of the globe for me), but am hoping that recordings of the sessions will be
made available.

-----Original Message-----
From: BlindMath [mailto:blindmath-bounces at nfbnet.org] On Behalf Of Godfrey,
Jonathan via BlindMath
Sent: Monday, June 26, 2017 5:58 AM
To: Blind Math list for those interested in mathematics
<blindmath at nfbnet.org>
Cc: Godfrey, Jonathan <A.J.Godfrey at massey.ac.nz>
Subject: Re: [BlindMath] Is it possible for screen reader users to do data
science?

Hello Dickson et al.,

I agree with Vincent. I would note that the vast majority of data science is
not actually a visual exercise. Experts put the proportion of work being
done to collect and clean data at anywhere from half to 90%. Yes, some of
the cleaning is done by investigating unusual observations in graphs, but
there are so many non-visual ways of gaining such information.

I would observe that statisticians were doing work well before graphs were
being created and that there are plenty of somewhat forgotten ideas that
might prove useful for a blind person to substitute for creating a graph.
For example, I have two variables and want to know if they are linearly
related. A sighted person jumps straight for a scatter plot and eyeballs the
results to help them decide if the relationship can be explained by a
straight line. While they are dithering, I've fitted the straight line model
and I'm checking the residuals because that's what is going to happen anyway
if the sighted person decides the relationship is linear. If the
relationship wasn't linear, I've found that out by investigating the
residuals; OK, I'm doing that with another scatter plot, but at least my
scatter plot is going to exaggerate any nonlinearity that existed in the
original relationship. I've also see the R squared of the model to decide if
there is any relationship to want to question he linearity of as well. Then,
if there is something to get excited about, I make the graph for the poor
sighted audiences I must work with.

The problem I have is that the use of the all too readily available software
that creates awesome data visualisations gets used in introductory teaching
to the detriment of the old-fashioned stuff that generally gets pushed into
the higher level courses. I was once asked if there was room to create an
accessible intro stats course for blind people. I didn't think so at the
time, but I've toyed with the issue on and off for the last few years. The
trend I've observed is to emphasise what you can see from the current
display of data, as against what a graph can show in a theoretical sense.
The same ideas get taught, but sort of in the wrong order for a blind
person. 

Compare the two quotes:
"Let's look at this data, and hey, what's that funny point over there? Well
it's called an outlier and this (insert name of graph here) can identify the
existence of outliers for you."
Versus:
"A (insert name of graph here) can highlight the existence of outliers.
Let's look at some examples, one with an outlier and  one without."

The first is disabling to anyone not able to see the graph in question,
while the second is less disabling. Unfortunately, the second (older) style
is considerably less stimulating for a sighted audience. The former goes
well with someone actually creating the graphs in a live demonstration while
the second suits the pre-prepared (I call it the cooking show approach)
style of lecturing. Actually most live demonstration style lectures require
some preparation too, but it's obvious that the actions are being completed
on the spot. Most smart lecturers now do the former and have the latter
(pre-prepared back up plan) waiting just in case.

Your choice of nonvisual technique will depend on the scenario in question,
and on your personal ability to comprehend what that nonvisual technique has
to offer.

I realise that some people swear by sonification as a substitute for visual
inspection of data.  I've also heard data sonification described as an
alternative to visual inspection of data with its use having an additive
impact to visual inspection. For others, sonification just makes a lot of
noise. Rather unfortunately, I find myself in this last category. I do like
that it can be done and I'll help anyone trying to create a useful tool for
sonifying data (in R only though) but that's because I'm interested in
getting a wide range of options for blind end users as against just working
towards tools that I personally will find useful.

I hope that you (Dickson) are considering coming to Orlando for the NFB
Convention where we will almost certainly have discussions on the problems
you face.

Cheers,
Jonathan

-----Original Message-----
From: BlindMath [mailto:blindmath-bounces at nfbnet.org] On Behalf Of Martin,
Vincent F via BlindMath
Sent: Monday, 26 June 2017 5:15 a.m.
To: Blind Math list for those interested in mathematics
Cc: Martin, Vincent F
Subject: Re: [BlindMath] Is it possible for screen reader users to do data
science?

It can be done using a number of methodologies.  I have done it with and
without sighted assistance.  Using a combination of tactile graphs and
sighted assistance is how I did forecasting work.  I also am using
sonification of data to interpret graphical output as well.  Work is being
done for sonifying graphs using "R", SAS has intergrated what we called the
"sonification sandbox" in Dr. Bruce Walker's sonification lab at Georgia
Tech and it is readily available right now.  It was unveiled at CSUN this
Spring and I use it with SAS right now.  I am also using sonified graphs
with the graphical output of SPSS in my current research as well.  I am up
to dissertation proposal on what type of experiment I want to do, but I am
sure it is similar to what most data analysts also use to make inferences
with their visual graphs as well.

-----Original Message-----
From: BlindMath [mailto:blindmath-bounces at nfbnet.org] On Behalf Of Dickson
Tan via BlindMath
Sent: Sunday, June 25, 2017 12:40 PM
To: blindmath at nfbnet.org
Cc: Dickson Tan <dickson.tan.2013 at gmail.com>
Subject: [BlindMath] Is it possible for screen reader users to do data
science?

Hi,

Given that visualization is used often in data science to see patterns, is
it possible for a screen reader user to do data science? I've heard of some
tools for doing basic bar/line graphing, but as far as I know, they are
immature, and can't handle other types of graphs. 

If not, how far would I get without being able to access visualizations?

Best regards,

Dickson

_______________________________________________
BlindMath mailing list
BlindMath at nfbnet.org
http://nfbnet.org/mailman/listinfo/blindmath_nfbnet.org
To unsubscribe, change your list options or get your account info for
BlindMath:
http://nfbnet.org/mailman/options/blindmath_nfbnet.org/vincent.martin%40gate
ch.edu
BlindMath Gems can be found at
<http://www.blindscience.org/blindmath-gems-home>

_______________________________________________
BlindMath mailing list
BlindMath at nfbnet.org
http://nfbnet.org/mailman/listinfo/blindmath_nfbnet.org
To unsubscribe, change your list options or get your account info for
BlindMath:
http://nfbnet.org/mailman/options/blindmath_nfbnet.org/a.j.godfrey%40massey.
ac.nz
BlindMath Gems can be found at
<http://www.blindscience.org/blindmath-gems-home>

_______________________________________________
BlindMath mailing list
BlindMath at nfbnet.org
http://nfbnet.org/mailman/listinfo/blindmath_nfbnet.org
To unsubscribe, change your list options or get your account info for
BlindMath:
http://nfbnet.org/mailman/options/blindmath_nfbnet.org/dickson.tan.2013%40gm
ail.com
BlindMath Gems can be found at
<http://www.blindscience.org/blindmath-gems-home>