Recently IBM Watson demonstrated the effectiveness of a natural language question answering algorithm. Of course, beyond the game of Jeopardy, this problem becomes more difficult. The accumulated body of mathematical knowledge seems to be a good target for domain specific natural language question answering algorithms, something like an automated Math Overflow. Is this being pursued?

13$\begingroup$ This is not really a math question. $\endgroup$ – Mariano SuárezÁlvarez Oct 4 '11 at 15:57

2$\begingroup$ The closest thing might be Wolfram Alpha. You can type any integral or sum and it will try to interpret what you gave it. You can even write things like "what is the monster group" and get a reasonable answer. $\endgroup$ – Eric Naslund Oct 4 '11 at 16:06

3$\begingroup$ I am voting to reopen, because I think that this is an interesting question that is potentially important for the practice of research in mathematics. Moreover, if anyone is in fact doing work of this type, then an unambiguously correct answer can be given by linking to that work; so the question is not merely an invitation for discussion. It should be community wiki, however. $\endgroup$ – Neil Strickland Oct 4 '11 at 16:48

3$\begingroup$ I agree that this is not a math question per se, but there are a lot of other questions on MO with a similar flavour. This is really a metamathematical question, which used to be considered 'mathematics' in the early part of last century. Just because this is not the kind of 'mathematics' most mathematicians worry about, is that really a good reason to close this? At least suggest where this question should be asked! $\endgroup$ – Jacques Carette Oct 4 '11 at 16:50

4$\begingroup$ While the application mentioned by the question is certainly very important, the question itself is much more of an "artificialintelligence" nature, not really a precise, clearcut math question. At best, it belongs to meta.MO, though more naturally it belongs to a machine learning type website. $\endgroup$ – Suvrit Oct 4 '11 at 17:10
I do not know whether there is anyone pursuing this, but I believe that there should be. All LaTeX files on the arxiv are publicly visible, and of course they contain richly structured markup with crossreferences between different parts of the same document and links to other documents. MathSciNet and Zentralblatt also have a lot of richly structured data. I think that a Watsonlike system could digest enough of this to give much better answers to certain types of questions than current search engines can. I don't think you could expect to provide actual answers to mathematical questions, but I think you could do a good job of pointing people to papers where answers could be found.

10$\begingroup$ I see that you do not look very often at the source of LaTeX files in arXiv! :) $\endgroup$ – Mariano SuárezÁlvarez Oct 4 '11 at 16:30

$\begingroup$ How do we view Latex source files on the arXiv? $\endgroup$ – Will Jagy Oct 4 '11 at 18:48

$\begingroup$ Alright, I did it, the relevant item was just two files. So, although they try to send as a .tar.gz , in this case i was able to just copy over to .tex and .bbl files with my own names. My computer insisted it was an archive and tried to manage things somehow... $\endgroup$ – Will Jagy Oct 4 '11 at 19:22

$\begingroup$ @Will: if you look at a page like arxiv.org/abs/1110.0405v1 You'll see a link marked "other formats" under "downloads" on the right. $\endgroup$ – Neil Strickland Oct 4 '11 at 19:33