MathCAT and the future of MathML


Presented at 2:15pm in Matchless on Thursday, November 17, 2022.



  • Neil Soiffer, Talking Cat Software

Session Details

  • Length of Session: 1-hr
  • Format: Lecture
  • Expertise Level: Beginner
  • Type of session: General Conference


MathCAT is an open source library that translates MathML to speech and braille. The talk demonstrates what is possible for accessible math. E.g., synchronized speech highlighting, prosody, advanced navigation, and support for both Nemeth and UEB on braille displays. Barriers that prevent greater accessibility with existing AT will also be discussed. The talk concludes with how some of the barriers are removed in a proposed MathML 4 that allows authors to guide the speech that should be used.


Accessible documents containing math typically make use of MathML, the standard for math on the web. JAWS, NVDA, Orca, VoiceOver/Safari, and MathJax/SRE are all capable of speaking MathML, navigating it, and converting it to Nemeth braille. This access is not limited to web documents: Microsoft Word and PowerPoint documents with math are also accessible because both MathType and the built-in Word editor can generate MathML.

For math speech, most people seem to prefer a semantic reading of an expression over a syntactic one (“sine of x squared” rather than “sin open paren x superscript 2 end superscript close paren”). To produce high-quality speech requires lots of rules or trained AI, and that is where speech differences tend to show up. Even with a large number of rules, there are cases where it is very difficult to tell what should be spoken because the meaning is very difficult to discern without help from the author. For example, “(3,6)” could be a point, an open interval, or even the GCD of two numbers. To disambiguate these cases and also to simplify the burden on AT speaking math, the W3C Math Working Group is developing ideas for MathML 4 that allow authors/authoring tools to express how an expression should be spoken.

MathCAT (Mathematics Capable Assistive Technology) takes advantage of MathML that includes the proposed mechanism in MathML for expressing author intent. It also has a large number of rules to try and infer intent for documents/expressions where it is not included in the MathML.

As is the case with speech, the quality of Nemeth generation in screen readers varies with most output being poor. MathCAT has focused on quality braille generation via over 700 tests (and counting) for the Nemeth and UEB technical braille it produces.

While MathCAT is meant to be a library that can be used by an AT, it specifically is meant to replace MathPlayer which has been discontinued. As such, an NVDA addon that can replace MathPlayer will be demonstrated.


  1. What math AT features are possible and what different screens provide so you can decide what is best for you
  2. MathCAT is a powerful library that converts MathML to speech and braille and can replace MathPlayer
  3. Latest developments regarding MathML 4 and how it improves accessibility by empowering authors/authoring tools

Disability Areas


Topic Areas

Assistive Technology, Research, Uncategorized

Speaker Bio(s)

Neil Soiffer

Neil Soiffer is a principal architect of MathML, the standard for putting math on the web. He was the main developer of MathPlayer, which is used with NVDA to make math accessible in Web, Word, and PowerPoint documents. He has published numerous papers on math accessibility and is a member of various standards groups concerned with accessibility on the Web and elsewhere. He currently chairs the W3C Math Working Group that is working on greater browser support for MathML along with updating the MathML standard. He received a B.S. in Math from MIT and a Ph.D. in Computer Science from UC Berkeley. He has worked at Tektronix's Computer Research Lab, Wolfram Research (Mathematica), Design Science (MathType, MathPlayer), and has now formed his own accessibility company, Talking Cat Software where he been focused on the open source MathCAT (Mathematics Capable Assistive Technology) project.


Slides for MathCAT and The future of MathML