computer-go: Need help building Graded Gnu Go Problem Database

["Pierce T. Wetter III" <pierce@xxxxxxxxxxxxxxxxxx>]

From: Pierce Wetter <pierce@xxxxxxxxxxxxxxxxx>
Date: Sat Jul 28, 2001  09:04:54 AM America/Phoenix
To: computer-go@xxxxxxxxxxxxxxxxx
Cc: gnugo@xxxxxxxxxxxxxxxxx, T.Wolf@xxxxxxxxxxxxxxxxx
Subject: Need help building Graded Gnu Go Problem Database

A while ago, there was a discussion about Graded Go Problems for 
Beginners....

The problem with using GGPFB is that the problems are copyrighted...

So I asked Thomas Wolf the author of GoTools if he would be willing to 
make a small fraction of his 24,000 problem set available as part of a 
public domain benchmark for go program developers.

He agreed! He's willing to make somewhere between 500-1000 problems 
available. Which is really cool, as the computer-go world will be able 
to have a standard set of problems they can exchange. I'll put them 
under the LGPL (Library Gnu Public License or "Lesser" GPL) so its clear 
that by including them your program doesn't be come public too...

That was a month ago (I took time off to learn perl to parse his problem 
database).

So yesterday, I worked out a program to convert from his format to sgf 
format. (in Perl!)

Now here's the rub: I need to now filter this problem set down to size 
(from 24,000 problems down to 500-1000 problems). I've also found that 
SGF isn't the best format for problem sets.

  So I need a combination of help and advice:

       I need help with the filtering and grading of the problems. 
You'll need access to your own go program for the grading step. You'll 
also need access to the database of problems, either by owning GoTools 
(preferred), or by having sent Thomas a license agreement for the 40,000 
problem set he distributes for non-profit use: 
http://alpha.qmw.ac.uk/~ugah006/gotools/t.wolf.gotools.problems.html  . 
Then I can send you versions of the perl program that converts the files 
to sgf format, and you can try them against your program... (That is, 
I'm distributing the program, not the problems. While we can distribute 
the resulting problem set freely, during the beta period, we'll only be 
distributing the programs...)

      For the filtering step, a program that could tell how similar two 
problems are would be useful, which requires much less go knowledge. 
That way the 500 problems we choose would be diverse as possible.

       I need advice on the best way to translate the information from 
GoTools format to SGF since SGF is really designed for games, not 
problem sets. Here you need knowledge of SGF in practice, as opposed to 
the spec.


  Three emails to follow, one on each of these topics, plus one 
"background" email on the problems.

Pierce

P.S.

  Thanks again to Thomas Wolf for allowing me to put this together!