Tuesday, December 20, 2011

The experience of reading a paper

Trying to read a proof in depth (say, because I was preparing a blog post on it), the reader may go through various feelings in sequence.

1. Stupidity; How come I don't understand it at all? Am I becoming stupid?
2. Anger and frustration; how come the authors leave some definitions ambiguous, and why did the referee not send the paper back to ask for another round of polishing?
3. An idea: let me just look for slides of a talk on the subject! Slides usually put much more emphasis on intuition and might help me understand what the authors meant to say.
4. Obstination; I can do it. There are no slides available, but I'm interested in the result and just need to put my mind to it. Go through examples systematically, try out the various interpretations until one fits and the proof goes through.
5. Suspicion. Is that result really correct? None of my interpretations seem to work.
6. Discouragement; I can't do it. I'll have to look up references of prior work and see which ideas are present in there and whether they are explained any better. This is turning into a major time sink.

As an intellectual activity, reading technical text is really not much different from doing research.


  1. How is it that there is no email to the authors as a step? Or would you like to do it all in a few hours?

  2. Email to the authors is the solution of last resort.

    They're usually not interested (who wants to go back to a paper from years ago and figure out if some stupid typo somewhere got in the reader's way?)

    In addition, it exposes my denseness. I should be able to gracefully jump over minor omissions and guess what the authors may have tried to imply without saying it explicitly.

    In addition, I have the feeling that the paper might be very readable to those who have followed the subject and have read the previous three or four papers on the subject. I am not in a rush to expose my lack of culture.

    What I might try to do, before that last resort step is discuss the paper with someone else who could provide a different perspective and maybe show me something simple that I might have missed or misunderstood.

  3. #7: Give up and read CS Professor's Blog.

  4. "As an intellectual activity, reading technical text is really not much different from doing research."

    This is sad. Reading should be in P whereas doing should be NP-hard.

  5. 7. Oh, screw it, I'll just prove it myself.

    8. Profit.


Note: Only a member of this blog may post a comment.