Marcin, in reaction to your quote below, I can only agree that the computational part is quite complex.
There was some research in Tilburg, resulting in a quite simple computational algorithm: - reduce all characters to their simple form (drop accents) - get their ascii value, raise it to the 5th power - add these numbers Result is a simple number to calculate with. All words with this number are letter-equivalent (or almost). - compute common letter group mistakes (hunspell rep) into number differences beforehand - apply a number of these differences to the main number - get their words Then add de character-distance cacluation for filtering out too different suggestions. This works rather well, as simplistic as it is. It does not take any compounding into account, which would result in less flexibility in suggestions for compounding languages. There shoul be a publication about tis algorithm from Martin Reynaert, University of Tilburg. Just in case you are interested. Ruud >>I'm not a fan of hunspell; I think it has a wrong approach for creating >>suggestions because the computational complexity of its algorithm is >>simply too high. ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Languagetool-devel mailing list Languagetool-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/languagetool-devel
