Dear Araki, actually I just wrote a master thesis on a the issue of concept learning on patterns.
I labeled the connectivity between two chains according to five classes: 1.strongly connected 2.connected (two moves can separate) 3.conditionally connected 4.separated (two moves can connect) 5.strongly separated With that I learnt a classifier using low level features from the Common Fate Graph defined by Thore Graepel et al. Results are quiet encouraging. My advice, only use annotations that you can record using first order logic and make the annotations specific. Try thinking of learning concepts that you believe are hidden inside your representation. This almost always means that you should try to learn local concepts. For instance you could learn the weakness of a group by labeling them from 1 to 10. The defence is due tomorrow, if the thesis is available by internet I led you know. Cheers, Emil Nijhuis On 12/14/06, Chrilly <[EMAIL PROTECTED]> wrote:
Le jeudi 14 décembre 2006 10:36, Stuart A. Yeates a écrit : > If I understand correctly, the point was that: > (a) parsing English is hard > (b) most English language comments on Go games are made by those for whom > English is a second language, who don't use "correct" English > :. (c) (b) is likely to make (a) even harder. > > Personally I disagree, but that's entirely off topic. > > cheers I think that (b) makes (a) much easier. English is very irregular language, and very comon mistakes are to "regularize" it (ed for past, "more" everywhere instead of "er"...) My personal experience is: i understand easyly people for whom english is not mother tongue, but i have bigger problems with native speakers. Yes, I worked at the European Space Agengy. The official language was English. The only ones which spoke a disturbing and difficult to understand language were the staff members from the U.K. Actually the official language was a sort of Pidgin. But this is also true for international conference papers and other sorts of international communication. I think it is on the one side nice to be able to communicate e.g. with people from Finnland without speaking Finnish. But after some time I found this Pidgin rather depressing. One can not communicate more subtle thoughts or feelings. Thats not only a language question, but also of the cultural background/semantics. Its e.g. also difficult for an Austrian to have a subtile conversation with a German, although the language is almost the same. But the German takes everything face value, the meaning of an Austrian sentence is often the opposite. E.g. "I enjoyed the party very much, it was very, very nice" means for an Austrian "It was a boring evening". btw, there was some times ago a disccussion about sgf format, and some ideas from PGN chess format: this one includes standardised annotation (http://www.very-best.de/pgn-spec.htm) ! good move !! very good move ? bad ?? very bad !? interesting move ?! dubious move +- white wins += white better =+ black better -+ black wins This is the Informator-standard in chess. There are additional characters like "with the idea of". Some of them are chess specific like e.g. the characters for pieces. this is easy to parse :) but not standard in go yet. Need some worldwide lobbying to convince chinese korean and japanese people to annotate games like this for ease of poor westerners guys, but i m rather sure they have some ideogram which says much more than this and will confuse us :) Or we are too lazy, ignorant to learn it. I have bought a "Japanese for Dummies" CD, but its too difficult for me. I admire the Japanese native speakers which learn English. I think it is also the other way round quite difficult. Chrilly _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
_______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
