2

I am looking for systems to make a research paper blind (i.e., remove all information about authors, affiliations, maybe even anonymize self-references to previous work) and do this automatically, i.e., without human intervention. Do such systems exist?

I have four main use cases in mind for this:

  • As a reviewer in non-double-blind venues, you may nevertheless prefer to review the papers of your batch by hiding information about the identity, affiliations, etc., of authors, to avoid biasing yourself. This probably wouldn't be easy in practice (as you often get emails about the papers, or see info about the papers on an editorial platform such as Easychair, which prominently feature the author information) but could be doable nevertheless (e.g., with the help of someone else to retrieve the papers for you). In this use case, the blinding system would take as input a PDF file of the article and hide author information, etc., probably using some PDF extraction tool such as Grobid.
  • As an editor of an overlay journal (where articles and submissions are hosted on a public repository), if you want your overlay journal to be double-blind, you will need to give reviewers a blinded version of the articles, but of course the original articles hosted on arXiv will have to feature the author information prominently, so you'll need to remove it. In this use case, for papers typeset in LaTeX, the blinding system could work from the LaTeX sources of the paper (distributed by arXiv), which is probably easier (but still tricky for non-anonymous self-references).
  • As an author, when (re)submitting your work to a double-blind conference, it would be helpful to automatically make the paper blind or at least have a way to flag potential violations of blindness (the most interesting ones being non-anonymous self-citations). Again this could work from the LaTeX sources.
  • Finally, when operating a double-blind conference, when authors submit their manuscript (as LaTeX sources or as PDF), it could be useful to automatically make it double-blind, or at least highlight possible double-blindness violations. (This is the previous point but on the conference side.)

[I am not sure whether this question is in scope here -- I hesitated with Software Recommendations but this is rather specific to academia so posting it here.]

1

No.

AI is getting smarter every day, but they're still not smart enough to realize that "In [paper] we showed [result]" is a self-reference that breaks anonymity. You will unfortunately have to anonymize papers yourself.

It's not a proof of course, but if such software existed, Elsevier would know about it. Since they didn't mention it in their "prepare for double-blind peer review" page, it doesn't exist.

1
  • Beyond such obvious mistakes, if the field is small enough you'll know either way. – orlp Mar 24 at 19:50

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.