Closed kostyfisik closed 8 years ago
For prep_on_end a good antipattern can be
token regexp [IVXLCM]*|[0-9]*
token в
token .
as soon as it if valid to put XX в. to the end of the sentence.
another antipattern is
token case_sensitive С|В|К|У
token .
(e.g. В. Г. Белинский)
For PNN_on_end - I believe that LT should not recommend aginst widely used patterns found in tatoeba base. I think that is can possibly be found few examples in tatoeba, then PNN on end is really a style error, however, this should be few cases, not hundreds.
Few PPN on end left
Only these rules are enabled: [PNN_ON_END] All spelling rules are disabled Working on: ../tatoeba-sentences-rus.csv Sentence limit: no limit Error limit: 100
Title:
Title:
Title:
Title:
Title:
Title:
Title:
Russian: 7 total matches
Russian: ø0.00 rule matches per sentence
I think the best solution is set default off this rule.
Switching off the rule by default is not a solution. It should be improved with an antipattern (or several antipatterns), the most obvious is to check the sentence for direct speech. I believe this can dramatically reduce the number of false positives and it can be again switch on by default.
Fixed.
They are rather simialr, like как и он/она как он/она чем он/она кто он/она кто же он/она or
VB:.*:Masc:.* он
VB:.*:Fem:.* она
tatoeba-result-ru-PNN_ON_END.txt