Wednesday, December 21, 2022

AsTeR: Spoken Math On The Emacspeak Audio Desktop

AsTeR: Spoken Math On The Emacspeak Audio Desktop

1. Dedication

To My Guiding Eyes

In fond memory of Aster who first showed the way for 10 years; to Hubbell and Tilden who ably followed her lead over the next 22+ years!

On The Internet
No one knows you're not a dog! Nor if you're the same dog! Or even the same gender!
Aster Labrador
 Hubbell Labrador
Tilden Labrador

2. Overview

The work describing Audio System For Technical Readings (AsTeR) was presented to the Faculty of the Graduate School of Cornell University on Monday, Jan 17, 1994 for my PhD. This release is an updated version of AsTeR that uses the Software DECTalk.

3. Implementation

  • AsTeR audio-formats TeX and LaTeX documents.
  • User interface is implemented in Emacs.
  • It uses the Emacspeak speech-server dtk-soft to connect to the software DECTalk.
  • The audio-formatter is implemented in Common Lisp (SBCL).
  • Emacs commands call Common Lisp via slime to communicate with Aster.

4. Prerequisites

  • Install Emacspeak 57.0 or later from Github.
  • Install slime and auctex using M-x package-install.
  • Install flex, SBCL and cl-asdf using the linux package manager.
  • Install Software DECTalk from Github.

5. Building AsTeR

  1. cd <emacspeak> to change to your emacspeak directory.
  2. Get source via git checkout
  3. cd aster-math/lisp && make

6. Usage

  • Add directory aster-math/ui/ to your Emacs load-path.
  • Run M-x load-library aster; M-x aster.
  • Aster commands are on Emacs prefix-key C-; SPC and C-' a.
  • M-x describe-function aster displays help.
  • To speak math using AsTeR:
    1. Send a TeX file.
    2. Send math content from any Emacs buffer.
    3. When editing LaTeX — including from within org-mode buffers.
    4. When browsing Wikipedia pages containing mathematics using Emacs' EWW browser. (Make sure to first disable shr-discard-aria-hidden).
    5. From Emacs Calculator (calc).
    6. From the Emacs interface to Sage — a symbolic algebra system.
    7. Papers from — see Accessibility Report
  • Once Aster starts speaking, you can use Aster's browser to move around.

7. References

  1. Demo recorded in October 2022.
  2. Demo recorded in 1994.
  3. Brian Hayes: Speaking Of Mathematics, American Scientist, March 1996 — An accessible overview of AsTeR.
  4. Envisioning Speech:Scientific American, Wayte Gibbs, September 1996 — Describes AsTeR, Audio-formatting and Emacspeak.
  5. Proceedings: RFB Math & Science Symposium, May 12 – 13 1994.
  6. PHd Thesis, January 1994.

Tuesday, December 20, 2022

Toward Accessible Scientific Documents From Arxiv.Org

Toward Accessible Scientific Documents Via Arxiv.Org

1. Background

I access research papers from multiple times a week. Having originally worked on access to STEM content for my PhD over 30 years ago, I find it both encouraging and challenging as I continue to acquire access to the research materials I need.

As the Digital Library of the future, arxiv is committed to improving the state of STEM Access; see the Accessibility report they recently published. A few months ago after talking to them about their goals, I wrote down some ideas that would make my own access to arxiv a smoother experience; posting them here so folks in the STEM Accessibility community can expand and build on these ideas.

2. Goal

Make Arxiv a destination for Accessible Scientific documents where blind students and researchers can consume the latest technical content via alternative modes of interaction that best suit their individual needs.

3. Making The Portal Easy To Use

A REST API to automate the download of multiple formats.

4. Making The Content Progressively Easy To Consume

  1. Build PDFs using PDFLaTeX rather than the DVI->PS->PDF pipeline.
  2. Ensure that PDFs are one-column layout.
  3. Build HTML+Math content from LaTeX using TeX4HT and friends.
  4. Build out a collection of high-quality LaTeX macros for specific sub-domains in CS and Math so authors dont need to invent their own marcos.
  5. Allow authors to contribute macros for new notation as it gets invented.
  6. Incorporate Speech Rules Engine (SRE) from Volker Sorge.
  7. Explore ChromeBooks where ChromeVox is already built-in.

5. Target Experience

Once we have:

  1. Arxiv: building PDF and HTML the way we desire.
  2. Client-side setup recipe – initially for a ChromeBook or Linux running my present environment (see final Teaser section).
  3. Mathjax and SRE injected at the right points on the backend at arxiv.

The target audience should be able to:

  1. Search and discover content on arxiv.
  2. View and consume content from Arxiv via the browser of choice with spoken feedback built-in.

6. Iterate And improve

  1. Test the flow with a few initial members of the target audience to discover the pain-points and fix them.
  2. Communicate what we learn via arxiv given the traffic and start publicizing the solution.
  3. Rinse and repeat ….

7. Research Areas

  1. Graphs – start with Graphviz — Dot Graphics.
  2. Graphs as in X-Y plots — there has been some work on data Sonification — but it's still early and exploratory.

8. References

  1. Speech Rule Engine from Volker Sorge.
  2. MathJax
  3. TeX4HT On Debian Linux, this is in package texlive-plain-generic.list — installed as /usr/bin/xhlatex.
  4. Emacspeak — includes support for Math via SRE — AKA poor man's AsTeR.
  5. Teaser: Heard from a bird: Speaking Of Mathematics: #emacspeak #STEM Again DECTalk → Again AsTeR.