Find file
a977288 Sep 23, 2016
329 lines (266 sloc) 13.9 KB
\conferenceinfo{ASSETS '16}{October 23-26, 2016, Reno, NV, USA}
\widowpenalty = 10000
\newcommand\ednote[1]{\typeout{There is still an editor's note!!!}%
\footnote{EDNOTE: #1}}
\newcommand\edbf[1]{\typeout{There is still an editor's note!!!}%
\textbf{EDNOTE: #1}}
% \DeclareUnicodeCharacter{B1}{\pm}
%% LANGUAGE Markup definitions
% commentstyle=\color{Brown},
columns=[l]flexible, %% , basewidth={0.5em,0.4em}
deletekeywords={type,id}, % does not work!
% otherkeywords={mfenced, open, close, separators, mrow, mi, mo, mn, math,
% msup, role, parent, children, added}
% \CopyrightYear{2016}
% \setcopyright{acmcopyright}
% \conferenceinfo{Assets,}{}
% \isbn{}}\acmPrice{\$15.00}
% \doi{}
\title{Supporting Visual Impaired Learners in Editing Mathematics}
%\subtitle{Extended Abstract}
% \alignauthor {T.V. Raman}\\
% \affaddr{Google, Inc.}\\
% \affaddr{Mountain View, CA}\\
% \email{\normalsize}
\alignauthor{Volker Sorge}\\
\affaddr{Progressive Accessibility Solutions, Ltd.}\\
\affaddr{University of Birmingham, UK}\\
%% Thanks does not seem to work.
We present an extension to the Emacspeak audio desktop that provides support
for editing mathematics in {\LaTeX}. It integrates MathJax and the Speech Rule
Engine to support users in reading {\LaTeX} sources, manipulating mathematical
formulas and verifying correctness of the rendered output.
\keywords{STEM Accessibility, Mathematics, MathJax}
Mathematics is still the single most significant hurdle for inclusive education
for visually impaired students in the STEM subjects. In particular, in secondary
education the increasing complexity of mathematical formulas makes it harder to
communicate content equally effectively to both sighted and blind students.
While mathematical formulas can be expressed in dedicated Braille formats, these
become quickly unwieldy for more advanced mathematics and moreover it becomes
impossible for students to communicate them to their peers or
professors. Consequently they need to start using a linear format that is
commonly understood. And indeed on transitioning from high school to University
students often have to learn {\LaTeX} to have a means to read and communicate
advanced material.
While there are some specialist editors that support editing mathematics for the
blind (e.g.~\cite{Pearson}) there is very little support for reading or
authoring {\LaTeX} documents directly (cf.~\cite{ncbys}). In particular, it is
difficult for students to write mathematical content in {\LaTeX} while checking
that the rendered output indeed corresponds to the mathematical formula they
intended to write. We aim to support this task by extending the Emacspeak audio
desktop~\cite{Emacspeak} with a mathematics option. It exploits the power of
MathJax~\cite{MathJax2.6} to translate {\LaTeX} expressions into MathML markup
and the Speech Rule Engine (SRE)~\cite{SRE} to translate MathML into speech
strings that can be passed to a TTS.
Our work focuses on supporting students in the aspects of learning, writing and
working with {\LaTeX}. In particular, to allow them to take source material,
read it, browse it and manipulate it. We support learning {\LaTeX} by offering
ways to write and rearrange expressions and hear the effect, by using MathJax's
error reporting mechanism to indicate incorrect expressions and by enabling
interactive exploration of rendered expressions on the fly, while editing. As
editing mathematics is already a time consuming task we aim to minimise the
impact of having to listen to lengthy expressions multiple times by exploiting
Emacspeak's prosody feature and SRE's ability to summarise sub-expressions
meaningfully for concise aural rendering of expressions.
\section{Technical Realisation}
Technically we realise our aims by combining Emacspeak with MathJax and the
Speech Rule Engine.
\textbf{Emacspeak} is a speech interface that allows visually impaired users to
employ the Emacs editor as the main tool for desktop activities. It supports
tasks from implementing in different programming languages, writing documents in
markup formats like markdown or {\LaTeX}, to browsing the file system and the
WWW. Emacspeak uses audio formatting via W3C's Aural CSS (ACSS) to produce rich
aural presentations of electronic information with the aim of shifting some of
the burden of listening from the cognitive to the perceptual domain. For
instance, syntax highlighting is aurally rendered by changes in voice
characteristic and inflection. Combined with appropriate use of non-speech
auditory icons this creates the equivalent of spatial layout, fonts, and
graphical icons important in a visual interface.
\textbf{Speech Rule Engine} (\sre) is a Javascript library that translates XML
expressions into speech strings according to rules that can be specified in an
extended Xpath syntax. It was originally designed for translation of MathML and
MathJax DOM elements for Google's ChromeVox screen reader~\cite{Sorge14}. It has
since been turned into a stand-alone application that can be installed server
side as a NodeJS application as well as used in client side web applications.
Besides the rules originally designed for the use in ChromeVox, it also has an
implemententation of the full set of Mathspeak rules. In addition it contains a
library for semantic interpretation and enrichment of MathML expressions, rules
for intelligent summarisation of sub-expressions, facilities for the interactive
exploration of mathematical expressions, highlighting of DOM nodes, etc. SRE
runs both in browsers as well as in NodeJS. It can be installed via Node's
package manager npm and offers a command line interface for batch translation of
XML expressions.
\textbf{MathJax} is a JavaScript library for visual rendering of formulas across
all platforms and browsers. It was originally built as a polyfill solution for
rendering mathematics in browsers until the MathML standard would be
sufficiently supported. But since most browsers still do not support MathML
natively, MathJax has become the quasi standard for displaying Mathematics on
the web. MathJax can render the most common mathematical authoring formats ----
{\latex}, ASCIIMath, and MathML --- into high quality output represented either
as HTML or as SVG. Although MathJax is built as a client side solution, to be
included into webpages, there is a special NodeJS package release, which allows
server side rendering of mathematical expressions from one of the three input
formats into SVG. In addition, it offers a way to convert {\latex} into
standardised MathML. MathJax can also use \sre as a dependency to produce speech
strings for mathematical expression for inclusion into the rendered output.
\textbf{NodeJS Bridge:} The three systems are combined with a simple NodeJS
bridge, whereby MathJax and \sre run within an inferior JavaScript process in
Emacs. Emacspeak communicates with both systems by message passing
between the different Emacs processes. Using an inferior process over a simple
batch process integration does not only decrease potential lag due to delay in
system startup but also has the advantage that both integrated systems can keep
an independent internal state Emacspeak can refer to.
\section{Reading {\LaTeX} in Emacspeak}
{\LaTeX} formulas are spoken in Emacspeak in two different ways: Either
integrated into the continuous flow when the entire document is read or
speaking can be triggered interactively for a single formula. In either case
Emacspeak detects the math delimiters and sends the expression to the node
process. MathJax then translates it into MathML and passes it to \sre, which in
turn performs a semantic analysis of the expression and then translates the
expression into a speech string using a chosen set of speech rules.
% \paragraph{Audio Formatting}
One of Emacspeak's features is that it enables audio formatting, by employing
changes in prosody (e.g., changes in pitch or stress), to indicate syntax
highlighting, special text elements like links etc. Similarly, \sre supports
prosody markup by allowing speech rules to define changes in pitch, rate, volume
and by adding pauses. We combine this by translating mathematical expressions in
\sre using a speech rule set that indicates two dimensional layout, such as sub
and superscripts or positions in fractions with changes in pitch and rate,
rather then employing more verbose indicator words like StartFraction or
EndFraction. \sre then returns translated formulas as sexpressions that contain
speech strings together with Aural CSS markup that Emacspeak can exploit for
audio formatting.
% \paragraph{Interactive Exploration}
Since mathematical formulas are often of a complex nature and just listening to
an expression once is generally not enough for a user to fully take in its
meaning, it is particularly important to be able to engage with them
interactively by offering a means of traversing sub-expressions. \sre provides
a number o different navigation options for mathematical expressions, which we
exploit from Emacspeak. Thereby \sre keeps an internal state on the latest
mathematical formula that it has rendered. Emacspeak then allows users to walk
this formula using simple cursor key navigation, while \sre produces speech
strings for the corresponding sub-expressions, potentially together with
positional information, such as denominator, numerator etc.
\section{Learning Support}
One major difficulty presented to students new to {\LaTeX} is learning the
intricacy of its syntax and in particular to correctly implement complex, nested
mathematical expressions. We provide a number of features that are aimed in
particular at helping students understand expressions and write correct ones
themselves. For this we exploit as much as possible the abilities of the systems
we integrate:
\textbf{Error Handling} Since MathJax implements a version of the TeX
typesetting system it can return fairly detailed error reports. We expose these
directly to the reader together with an error earcon to allow users to quickly
understand and rectify problems in their input expressions.
\textbf{Syntax Highlighting} Since {\LaTeX} documents are implemented rather
than composed in a WYSYG style, editors help authors to distinguish content from
markup using syntax highlighting. Emacspeak can exploit its audio formatting to
achieve a similar effect, making it possible to separate content from markup and
giving the reader an indication of what layout elements markup would produce,
such as bold type face, section headings etc. For math expressions audio
formatting is used to indicate nesting depth of braces, giving a user an
indication of the position within an expression they edit thus highlighting
potential problems or errors.
\textbf{Minimise Aural Rendering} Since editing {\LaTeX} can be a time consuming
task, which is only compounded by the need to listen to lengthy audio output of
mathematics, one of our main goals is to decrease the burden of listening. One
way is by prefering rich audio formatting to more verbose traditional reading
styles of mathematics (cf.~MathSpeak for example). A second way is by exploiting
the summarisation features \sre provides. For example, instead of reading a
complete expression, \sre can use its semantic interpretation to produce
summaries such as ``sum with 5 summands'', which are only expanded when
explicitly prompted by a user, thus shortening aural output while editing.
% Emacspeak uses this during editing of single
% expressions.
% When the same expression is rendered multiple times, \sre can keep
% an internal state of the expression allowing it to simply summarise
% sub-expressions that have not been changed since the last editing state.
The author is grateful for TV Raman's technical work on the Emacspeak
MathJax Consortium.
\newblock {MathJax} v2.6, 2015.
\newblock \url{}.
S.~Dooley, D.~Brown, E.~Lozano, S.~Park, and S.~Osterhaus.
\newblock Online nemeth braille input/output using content mathml.
\newblock In {\em Proc of W4A}. ACM, 2016.
\newblock Blindmath gems {LaTeX}.
\newblock Emacspeak.
\newblock \url{}.
\newblock {S}peech {R}ule {E}ngine.
V.~Sorge, C.~Chen, TV~Raman, and D.~Tseng.
\newblock Towards making mathematics a first class citizen in general screen
\newblock In {\em Proc of W4A}. ACM, 2014.
%%% Local Variables:
%%% mode: latex
%%% TeX-master: t
%%% End: