MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am in the process of scanning a large collection of handwritten notes. They consist of diagrams and formulae with a relatively small proportion of actual words. Of course it would be hopeless to get an OCR program to digest the diagrams or formulae, but it would be useful if I could get one to find and transcribe enough of the words to build an index. Has anyone tried this kind of thing?

share|cite|improve this question
Tangentially related:… (It is not entirely hopeless to get OCR to digest formulae.) – Willie Wong Apr 30 '12 at 8:08
Use wetware. It needn't even be grad students: ordinary MSc students work fine for this. They will transcribe most of the formulas too, though usually not the diagrams. – Zsbán Ambrus Apr 30 '12 at 9:21
I thought MSc students are also grad students, by definition. Anyway, if you do opt for "wetware", be sparing... :) – Felix Goldberg Apr 30 '12 at 11:29
The question is specific to mathematics because OCR systems use heuristics that may well break if the words are interspersed between non-words. If many people had used many different systems then they would probably have information about which systems were more or less affected by this issue. But perhaps the relevant experiments have not been done yet. – Neil Strickland Apr 30 '12 at 13:47
This is much more suited to stack overflow, voting to close. – Igor Rivin Apr 30 '12 at 17:49

I recommend They have a trial version, with long enough trial period to do a big project. I've used InftyReader on a flakey old laptop and it did a pretty good job for a math book reissue. It took about half an hour or so to do each 40 page bundle of the disassembled book. It made systematic errors in the tex, many of which I was able to correct with some awk commands, but of course I still had to go through the whole thing tediously. I was not going for perfection, but just for something editable, and I think it came out better than required for this purpose.

share|cite|improve this answer
The notes are handwritten I assume the book was not. – user9072 May 1 '12 at 1:09
While I am very happy to have learned about the existence of InftyReader, I can confirm that it was not able to digest my handwritten notes. – Neil Strickland May 1 '12 at 7:43

To supplement Bob Terrell's post, here is an example from InftyReader.

Snippet from an input image, scanned at 600dpi:

Snippet from output of the corresponding LaTeX produced by InftyReader:
  LaTeX Output
It's not perfect—$\partial z \partial \bar{z}$ becomes $\partial z k$—but it's pretty impressive!

share|cite|improve this answer
Did you try it with something handwritten, too? This iw what the question seems to be about. – user9072 May 1 '12 at 1:11
@quid: No. I am less interested in the question Neil posed than I am in learning what this neat software can do. InftyReader is designed to work on scanned images of printed mathematics, not handwriting. – Joseph O'Rourke May 1 '12 at 1:44

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.