Hi, W dniu 2013-07-03 00:03, Jaume Ortolà i Font pisze: > Hi, > > You can see what has happened in the Wikipedia checks. See the links below. > > In some languages, there are false alarms removed: French, Breton and > Catalan. That looks good. > Other languages have added alarms: English, German, Russian, Polish and > Italian. The reason is that these languages had previously a special > treatment that has been now removed.
And all alarms are false. No false alarm has been removed. I think this is a clear case of a regression. > > The question is what to do with sentences that end with no ending > punctuation mark (.?!...). If we don't require uppercase sentence start > in these sentences, we avoid a lot of false alarms in lists, tables, > etc., as you can see in the Wikipedia check. On the other hand, we can > get false negatives, as in the reported bug, in titles, etc., when (by > mistake or not) there is no punctuation mark at the sentence end. Polish has special norms about titles: you can use a question mark but not a dot at the end (even if it's a complete sentence). > > We can try a midway solution: don't require upper case sentence start > when both the previous and the current sentence have no ending > punctuation mark. This situation is what we can find in a list or a > table, and we can surmise it isn't an accumulation of mistakes. > > What do you think? Any ideas? Well, I'm not sure but your current solution definitely does not work for Polish. Regards, Marcin > > Regards, > Jaume Ortolà > > > LanguageTool Nightly Diff Overview 2013-07-02 22:20 > > This page lists the results of our automatic nightly testing against a > fixed Wikipedia corpus with 1000 articles per language. > > Changes 2013-07-01 22:20 to 2013-07-02 22:20 > Version: 2.3-SNAPSHOT (2013-07-02 22:02) > [1]Changed: en > [2]Changed: de > [3]Changed: fr > [4]Changed: ru > [5]Changed: br > [6]Changed: ca > [7]Changed: pl > [8]Changed: it > > Total runtime: 2013-07-02 22:20 to 2013-07-02 23:10 > > References > > 1. > http://languagetool.org/regression-tests/20130702/result_en_20130702.html > <http://languagetool.org/regression-tests/20130702/result_en_20130702.html> > 2. > http://languagetool.org/regression-tests/20130702/result_de_20130702.html > <http://languagetool.org/regression-tests/20130702/result_de_20130702.html> > 3. > http://languagetool.org/regression-tests/20130702/result_fr_20130702.html > <http://languagetool.org/regression-tests/20130702/result_fr_20130702.html> > 4. > http://languagetool.org/regression-tests/20130702/result_ru_20130702.html > <http://languagetool.org/regression-tests/20130702/result_ru_20130702.html> > 5. > http://languagetool.org/regression-tests/20130702/result_br_20130702.html > <http://languagetool.org/regression-tests/20130702/result_br_20130702.html> > 6. > http://languagetool.org/regression-tests/20130702/result_ca_20130702.html > <http://languagetool.org/regression-tests/20130702/result_ca_20130702.html> > 7. > http://languagetool.org/regression-tests/20130702/result_pl_20130702.html > <http://languagetool.org/regression-tests/20130702/result_pl_20130702.html> > 8. > http://languagetool.org/regression-tests/20130702/result_it_20130702.html > <http://languagetool.org/regression-tests/20130702/result_it_20130702.html> > > > 2013/7/2 Jaume Ortolà i Font <[email protected] > <mailto:[email protected]>> > > Hi, > > There is a bug report about the behavior of UppercaseSentenceStartRule: > > https://sourceforge.net/p/languagetool/bugs/185/ > <https://sourceforge.net/p/languagetool/bugs/185/> > > I think that the only situation in which we can safely prevent the > rule to match is when the previous sentence ends with comma or > semicolon. So I propose to implement this for all languages. > > Perhaps we can do the same when the previous sentence ends with no > punctuation mark at all. This could be useful for table cells, but > sometimes there will be ambiguities. I am not sure. > > The current implementation looks at the sentence end to decide what > to do at the start of the same sentence. I think this makes no sense > and causes false negatives. > > I can make some changes and we'll be able to see what happens in the > wikipedia checks. > > Regards, > Jaume Ortolà > > > > > ------------------------------------------------------------------------------ > This SF.net email is sponsored by Windows: > > Build for Windows Store. > > http://p.sf.net/sfu/windows-dev2dev > > > > _______________________________________________ > Languagetool-devel mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/languagetool-devel > ------------------------------------------------------------------------------ This SF.net email is sponsored by Windows: Build for Windows Store. http://p.sf.net/sfu/windows-dev2dev _______________________________________________ Languagetool-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/languagetool-devel
