+Mohan Ganesalingam and I have just posted a preprint on arXiv about a fully automatic theorem-proving program we wrote a few months ago and which I discussed on my blog. Also, the output of the program on our current set of test problems can be found here: http://people.ds.cam.ac.uk/mg262/robotone.pdf
We'd be grateful for any comments or criticisms, whether superficial or more substantial. (But perhaps I'll say in advance that we're aware that some of our urls go over the end of the line and that the top of the first page looks a bit odd. It was more trouble than it was worth to deal with those two glitches.)
#spnetwork arXiv:1309.4501
We'd be grateful for any comments or criticisms, whether superficial or more substantial. (But perhaps I'll say in advance that we're aware that some of our urls go over the end of the line and that the top of the first page looks a bit odd. It was more trouble than it was worth to deal with those two glitches.)
#spnetwork arXiv:1309.4501
View 31 previous comments
+Mohan Ganesalingam What programming language(s) did you decide to write the program in?Sep 20, 2013
It's in Haskell.Sep 20, 2013
I am currently reading 'The Haskell Road to Logic, Maths and Programming': http://homepages.cwi.nl/~jve/HR/
For anyone interested in learning to program, in Haskell, and how to write proofs in that language, it's a great book.Sep 20, 2013
All you need to get started is http://hackage.haskell.org/platform/ (very easy to install).
For a pure introduction to the Haskell programming language see this free online book: http://learnyouahaskell.com/Sep 20, 2013
Two more hints on related work:
- http://www.naproche.net/ Natural language processing combined with ATP.
- De Bruijn was a (the?) pioneer on the mathematical vernacular. Did you look at his work?
http://alexandria.tue.nl/repository/freearticles/610209.pdf
www.cs.ru.nl/~freek/notes/mv.pdf (comparison with Isar, Mizar, ...)
Also note de Bruijn's emphasize on typed sets. It's similar in spirit with Lawvere's set theory.Sep 20, 2013
+Bas Spitters Thanks for the references -- but Naproche, Automath, Mizar and Isar are all about human-like input, whereas we are producing human-like output.Sep 20, 2013