4

Mathpix can convert formulas in journal articles into Mathjax/LaTeX code by scanning or taking a picture of them.

Is there a similar way to scan a journal article from the 1970s that was written on an old-fashioned typewriter, either in its entirety or just its maths, so that the formulas written in typewriter are converted to LaTeX code? (Computer Modern font)

This is an example article written by manual typewriter that needs to be converted to LaTeX:

As you can see, formulas written by typewriter make it very difficult to follow any sort of derivation. Images in the article of course can be left as is.

Also, does automated conversion to LaTeX exist somehow at the document level? rather than equation-by-equation and page-by-page conversion

5
  • 1
    it looks like my thesis, it can't be that old fashioned:-) – David Carlisle Nov 13 '20 at 17:27
  • would there be a big "market" for refreshing publications "from back-then" – develarist Nov 14 '20 at 11:31
  • 1
  • are we allowed to refurbish academic articles and make the refurbished versions open source – develarist Nov 14 '20 at 11:57
  • copyright law is complicated and region specific, best to assume no, unless you have a lawyer, but see google books (but they have lots of lawyers) – David Carlisle Nov 14 '20 at 12:12
7

You mention mathpix, but have you tried it on the document?

I tried one of the formula from the link you gave:

enter image description here

mathpix produced

    \begin{equation}
x_{k}=\frac{E \Sigma_{1}^{m} v_{k j}\left(C E_{j}-A\right)+\sum_{1}^{m} v_{k j}\left(B-A E_{j}\right)}{D}, k=1, \cdots, m
\end{equation}

Which hand correcting one Sigma to sum and wrapping in a latex document produces this

enter image description here

\documentclass{article}

\begin{document}

\begin{equation}
x_{k}=\frac{E \sum_{1}^{m} v_{k j}\left(C E_{j}-A\right)+\sum_{1}^{m} v_{k j}\left(B-A E_{j}\right)}{D}, k=1, \cdots, m
\end{equation}
\end{document}

which given the rough format of the input is pretty good I think.


To complete the circle, if you add

\usepackage{typewriter}

and process with lualatex, the resulting output is:

enter image description here

\documentclass{article}

\usepackage{typewriter}
\def\sum{^^^^03a3}
\begin{document}

\begin{equation}
x_{k}=\frac{E \sum_{1}^{m} v_{k j}\left(C E_{j}-A\right)+\sum_{1}^{m} v_{k j}\left(B-A E_{j}\right)}{D}, k=1, \cdots, m
\end{equation}
\end{document}
5
  • does automated conversion to LaTeX exist somehow at the document level? rather than equation-by-equation and page-by-page conversion – develarist Nov 13 '20 at 17:43
  • also, without using Mathpix scanning, isn't there a typewriter to Math Computer Modern automated conversion already, forward as well as backward? this would be a more direct route, although I guess there would be no escaping the scanning step – develarist Nov 13 '20 at 17:45
  • @develarist mathpix can (in its latest version) OCR mixed text and equations try it and see how far you can push it. (it's free after all:-) – David Carlisle Nov 13 '20 at 17:57
  • @develarist why would you expect an automated translation already? mathpix does a pretty amazing job, especially as it is usable for free in most cases. what kind of "direct" route would you expect?, you need complicated heuristics to spot that that line of --- is a fraction bar and gather up the numerator and denominator, the typewriter or computer modern font issue is just a cosmetic trivial detail compared to recognising the math layout. – David Carlisle Nov 13 '20 at 17:59
  • @DavidCarlisle -- If I were really being fussy, I'd change that \cdots to \ldots or just \dots. (The translation of this fraction from typewriter to LaTeX is amazing!) – barbara beeton Nov 16 '20 at 2:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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