Smyslov_Fan wrote on 05/12/14 at 13:08:59:


Not really.

Let me repeat my suggestion of letting computers do the sorting. 

1) Come up with a good classification scheme. (Probably the hardest part)

2) Write a program which classifies fens/pgn according to that scheme by using the fens and either
    a) the initial positions 
    b) all positions   
contained in the pgns.

Such a program can be written using existing libraries very quickly if your classification scheme is based only on material balance. It is a bit harder if you want to include classifications such as Rfgh+a vs Rfgh or include stuff like existence of past pawns.

3) Run the program on inserted/attached fens/pgns to add the results to the posts automatically.
 
One reasonable option is to first classify by material and then decide later if the the classification scheme (and the sorting program) needs to be refined.

If you want to extract all the information (moves embedded in the text) then it would require more work but is still doable. 

TalJechin wrote on 05/12/14 at 10:10:24:


Yes, one or the other sounds good. As soon as we upgrade to the new version of the YaBB software we will be able to have sub-boards, so that will help classification a lot.

That classification is about the material. If I understand correctly it's about a classification adding knowledge about positions.

Smyslov_Fan wrote on 05/11/14 at 17:32:21:


Yes, that sounds like a good idea, but an awful lot of work!

We could classify the endgames by the material. That's fairly common, but I don't know how helpful it will be when searching for R+4 pawns vs R+3 pawns endgames with pawns all on one side, for example. That would require an index with links to individual posts more than just a classification for each thread. 

If you want a useful classification system, perhaps it would be helpful to create one from scratch and make it a separate part of the subject heading for each new poster to fill out? Given enough time, I could add those classifications to threads retroactively (but not individual posts).

GMTonyKosten wrote on 05/10/14 at 22:30:29:


Depends on the amount of work you want to put in. A fairly straightforward approach is to go through all endgame posts and look for fen's and attached pgn files and index all posts using material balance or some more elaborate ECE like scheme. Extracting analysis from text is possible but more serious work (you could get a database of all endgame positions discussed here that way).