TY - GEN
T1 - A Smooth Transition to Modern mathoid-based Math Rendering in Wikipedia with Automatic Visual Regression Testing
AU - Schubotz, Moritz
AU - Sexton, Alan
PY - 2016/7/22
Y1 - 2016/7/22
N2 - Pixelated images of mathematical formulae, which are inaccessible to screen readers and computer algebra systems, disappeared recently from Wikipedia. In this paper, we describe our efforts in maturing mathoid, the new services that provides better math rendering to Wikipedia, from a research prototype to a production service and a novel visual similarity image comparison tool designed for regression testing mathematical formulae rendering engines. Currently, updates to Math rendering engines that are used in production are infrequent. Due to their high complexity and large variety of special cases, developers are intimidated by the dangers involved in introducing new features and resolving non critical problems. Today’s hardware is capable of rendering large collections of mathematical contents in a reasonable amount of time. Thus, developers can run their new algorithms before using them in production. However, until now they could not identify the most significant changes in rendering due to the large data volume and necessity for human inspection of the results. The novel image comparison tool we are proposing, will help to identify critical changes in the images and thus lower the bar for improving production level mathematical rendering engines.
AB - Pixelated images of mathematical formulae, which are inaccessible to screen readers and computer algebra systems, disappeared recently from Wikipedia. In this paper, we describe our efforts in maturing mathoid, the new services that provides better math rendering to Wikipedia, from a research prototype to a production service and a novel visual similarity image comparison tool designed for regression testing mathematical formulae rendering engines. Currently, updates to Math rendering engines that are used in production are infrequent. Due to their high complexity and large variety of special cases, developers are intimidated by the dangers involved in introducing new features and resolving non critical problems. Today’s hardware is capable of rendering large collections of mathematical contents in a reasonable amount of time. Thus, developers can run their new algorithms before using them in production. However, until now they could not identify the most significant changes in rendering due to the large data volume and necessity for human inspection of the results. The novel image comparison tool we are proposing, will help to identify critical changes in the images and thus lower the bar for improving production level mathematical rendering engines.
M3 - Conference contribution
T3 - CEUR Workshop Proceedings
SP - 132
EP - 145
BT - Proceedings of the 9th Conference on Intelligent Computer Mathematics
PB - CEUR-WS.org
T2 - 9th Conference on Intelligent Computer Mathematics
Y2 - 25 July 2016 through 29 July 2016
ER -