I entered the press room at JMM this morning to find a bustle of excitement as four men prepared to introduce themselves on video. As they explained their roles to the camera, J. Brian Conrey of AIM pulled me aside to tell me about their project: translating math textbooks into Braille automatically, something that until now took six months and $10,000. I sat down and asked them a few questions.
Al Maneki, a retired mathematician who is blind, was an initial proponent of the project and tests the team’s results to make sure they actually work. He and Martha Siegel initially conceived of the idea, and recruited Alexei Kolesnikov of Towson University. When they hit a wall, they asked J. Brian Conrey and David Farmer (also of AIM) for help. A rough, edited—and incomplete despite my typing efforts—of my conversations follows.
David: Martha and Al and Alexei were working on this and hit a wall. Then we had a meeting with the right people—Volker Sorge and Rob Beezer—at the last JMM. They had the key pieces that were missing.
How did you know to who to introduce Al, Martha, and Alexei to?
David: We had a workshop about screen readers. I knew the people involved in that would be interested in this. There’s really two parts: the narrative part of a book which we take for granted because we can see a title and we know that’s the chapter title. But in Braille everything is the same size. Rob Beezer had built this infrastructure that could handle that. Volker Sorge had been involved in MathJax which puts math on webpages. It has a visual display for sighted people and a structural “pronounce-it” for blind people, and now it also outputs Braille—it talks to the Braille reader. There’s a third part—the images—which Alexei is taking the lead on and that’s the hardest part. Say you want to show a graph that has two curves on it so that blind people can feel it. The resolution of your fingers isn’t big, so if you want it to be in Braille it needs to be much bigger. The images are still an art and we’re having a workshop in August, organized by the team, to try to turn that into a science.
How did the project get started?
Alexei: My colleague, Martha Siegel, a professor emerita at Towson University and a longtime secretary at MAA, was particularly moved by a student who had to delay taking a math course for six months because the textbook wasn’t available. People had to take the source textbook and type it out in Braille. I thought this was a little ridiculous. I thought surely, in 2018, a better way would exist. I started looking at what is available, and nothing worked. We could produce raised print, but we would give it to Al to test, and there would be a lot of problems. Eventually we realized the direct path of converting LaTeX to raised print is pretty difficult because of the styling and so forth. So PreTeXt offered a different path from mathematical text.
What is PreTeXt?
Rob: PreTeXt is a source format that separates content from presentation. We know exactly where the math pieces are and where the sections and chapters begin. It’s totally automated. We now have about 100 projects in PreTeXt, covering almost all the undergraduate curriculum besides topology.
Is the problem of producing textbooks in Braille unique to math?
Al: It’s most difficult for math. Braille textbooks in English can be had if you have an electronic file. English can automatically be translated to Braille. When it comes to math with equations and graphics, tables and matrices, that’s the hard part.
Volker: It’s hard for STEM, generally.
Al: There’s a standardized way, Nemeth, for representing math, diagrams, tables, and charts in Braille.
Volker: That [Nemeth] used to be done by humans. Now you can do it mostly automatically.
Why has automating Nemeth not been done before?
Volker: It has been done before, but those systems don’t work with Latex as well. The system I’m using is that we actually restructure all the math into something that’s natural to speak. We put it in a more semantic representation.
Al: There are very complicated formulas too—subscripts inside exponents, etc. It’s clearer in Braille than when you listen to it spoken. That can be hard because you can’t go back to reference it. There’s an art to how you speak math, and if you don’t practice that art just so, then you get into ambiguities, whereas in Braille if something’s ambiguous you can reference it again.
Are the Braille textbooks being adopted anywhere yet?
Rob: We are past proof of concept, but it’s not being adopted anywhere yet.
Alexei: One challenge is that a 15-page pdf becomes a very long chapter in Braille (about 3 times longer). There is no ability to control the size of the font. And you have only forty characters to play with on a line. It’s also surprisingly difficult to make sure symbols are translated correctly, because some rules are rules, but some rules are only guidelines.
Rob: We will eventually be able to distribute an electronic version of this and people can read it from the web on their Braille reader.
Volker: And students will be able to emboss research papers for themselves in the future.