Towards Universal Rendering in MathJax
Davide Cervone, Peter Krautzberger, Volker Sorge · 2016 · Proceedings of the 13th International Web for All Conference (W4A) · doi:10.1145/2899475.2899494
Summary
This paper presents the MathJax Consortium's effort to transform MathJax from a purely visual mathematics rendering library into a universal rendering solution that supports readers with visual, print, and motor impairments. The authors describe how mathematical content on the web has long suffered from poor browser support for MathML, the dedicated markup language for mathematics. While MathJax solved the visual rendering problem by converting LaTeX, ASCIIMath, or MathML into consistent cross-browser displays using HTML/CSS or SVG, accessibility remained an unsolved challenge. Before this work, math accessibility on the web was largely dependent on specific platform and browser combinations, with MathPlayer on Internet Explorer being the primary solution until IE11 deprecated plugin support. The paper introduces a semantic enrichment process that rewrites flat Presentation MathML expressions into semantic term trees, embedding this richer structure as HTML5 data attributes within MathJax's rendered output. This approach ensures that the semantic information is preserved regardless of which visual renderer MathJax uses, providing a foundation for multiple accessibility features. The authors tested their solution across six major screen readers (NVDA, JAWS, WindowEyes, VoiceOver, ChromeVox, and Orca) on multiple operating systems and browsers, aiming for a consistent experience regardless of the user's particular assistive technology setup.
Key findings
The paper demonstrates five key accessibility features built on MathJax's semantic enrichment: (1) Highlighting that uses semantic structure rather than syntactic structure to provide meaningful color-based emphasis of sub-expressions, particularly beneficial for readers with dyslexia. (2) Structural abstraction that simplifies complex equations by collapsing sub-expressions into symbolic indicators (e.g., a collapsed sum shows as a plus icon), allowing users to progressively expand and explore formulas at their own pace. (3) Magnification that exploits semantic markup to perform line-breaks at mathematically appropriate positions, avoiding the readability problems of naive line-breaking. (4) Aural rendering using the MathSpeak rule set, with speech strings exposed to screen readers via ARIA live regions, making math voicing independent of the screen reader's own math capabilities. (5) Interactive exploration allowing keyboard-driven navigation through formula sub-expressions along either syntactic or semantic tree paths. The authors identify significant technical challenges including inconsistent ARIA implementations across browsers, incomplete integration with OS accessibility APIs, and the lack of a dedicated ARIA module for STEM semantics.
Relevance
This work represents a pivotal moment in making mathematical content accessible on the web, addressing a fundamental gap in STEM education and information access. For accessibility practitioners, it highlights how semantic enrichment of content can enable multiple accessibility modalities from a single source. The paper's call for a dedicated ARIA module for STEM content remains relevant, as mathematical accessibility continues to depend heavily on library-level solutions rather than native browser and assistive technology support. The identified challenges around cross-platform AT testing and inconsistent ARIA implementations reflect ongoing issues in the accessibility field. Organizations publishing mathematical or scientific content should consider MathJax's accessibility extensions as essential infrastructure rather than optional enhancements.
Tags: STEM accessibility · mathematics accessibility · MathML · screen readers · universal design · assistive technology · web accessibility
Standards referenced: MathML 3.0 · HTML5 · ARIA · CSS