Re: Solving Meta Sentences

To improve the algorithm, I need more sentence beginnings that I know have a resolution, for analysis. So please let me know of any successes you have.

As robin pointed out here a number of successful sentences can be found at this site.

Can you give us an idea of what kind of output we should expect? I ran your code (with my own starting string "I wish I could write a sentence using") and got lots of stuff like this:

2 94-acdefghinorstuvwxy
3 89-cefghinorstuvwxy
4 68-cefghinortuvwy
7 45-efghinorstv
8 26-efhilnorstuvwx
10 22-fghinorstuwxy
[download]

Will I get the full sentence if I wait long enough?

It's certainly an interesting puzzle.

jarich

Comment on Re: Solving Meta Sentences Download Code

Replies are listed 'Best First'.
Re: Re: Solving Meta Sentences by Molt (Chaplain) on May 30, 2002 at 09:58 UTC
I know this was covered in Hofstadter's book 'Godel, Escher, Bach' (Excellent book, well worth reading), and seem to recall that if you could prove that every sentence would eventually terminate, or prove that a particular sentence would never terminate, then you would have just proven a rather major mathematical theory. Bonus points for proving it with a Perl script :)	[reply]
Re: Re: Solving Meta Sentences by YuckFoo (Abbot) on May 30, 2002 at 06:52 UTC
This output lines are just to convince you that the program is running. The first number is the number of iterations. Next is the score of the best attempt so far followed by the letters whose count is not right. Score of 5 means the whole sentence is 'off by 5'. Every $RESET iterations the best number is reset so you can see more work being done. I don't know if eventually a sentence will resolve. I tend to think some sentences don't have a solution. I've run the program for many millions of iterations over many hours and not found an answer. Even with a run this long I'm sure only a small fraction of possibilities were tried. I've also had it finish in five minutes. I think I need to run the same sentence many times to see if there is any consistancy about how many iterations it takes. GoodLuckFoo	[reply]