On 2015-06-08 21:27, Jaume Ortolà i Font wrote: > You are right. These results are not expected. I will look at them > again.
I have another problem now with the Morfologik snapshot and release: "is" is a typo (German), the best suggestion would be "ist". But I also get "die" because "die" and "ist" both have a distance of 1 when I debug the CandidateData constructor. What kind of "distance" is that parameter supposed to be? An edit distance like Levenshtein or some simplification that ignores the fact that "is" is closer to "ist" than to "die"? I know the answer is somewhere in the Oflazer paper but I don't have time now to work through that... Here's a test case that shows the problem when used with a debugger: https://github.com/languagetool-org/languagetool/blob/master/languagetool-language-modules/de/src/test/java/org/languagetool/rules/de/GermanSpellerRuleTest.java#L297 Regards Daniel ------------------------------------------------------------------------------ _______________________________________________ Languagetool-devel mailing list Languagetool-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/languagetool-devel
