Re: [computer-go] SGC

[Don Dailey <drd@xxxxxxx>]

> When doing  machine learning  on go game  collections, I find  it most
> convenient  to use  a  "one-line  format" (.olf)  which  puts all  the
> relevant data of a game  (players, ranks, handicap, komi, score, etc.,
> and  moves)  on  a single  line  (Windows  users,  think .csv  file  -
> comma-separated  values).  Why?   There are  a lot  of free  Gnu tools
> (grep,  sed,  awk, perl,  etc.)  which allow  me  to  easily pull  out
> sub-sollections  of  games according  to  various  criteria from  .olf
> files.  Gives  me all the  features of a  database without any  of the
> hassle.

Yes, this  is exactly what I  do too.  When  I'm doing my own  stuff I
never make  multi-line records.   

Of course  it's easy to  convert most muli-line records  (depending on
how  messed  up  they are)  to  single  line  records using  the  same
utilities that you mentioned.

- Don

_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/