Back to LanguageTool Homepage - Privacy - Imprint

French 'accord en nombre' problem in v1.7


(Sylvain) #1

Hello,

It seems that the rule "pluriel suivi du singulier" does not work in the 1.7 version, it worked in earlier versions.
For exemple very simple mistake like :
"les faute", "des dessin" ... do not match.
Do you know whats happenig ?
Thanks

Sylvain


(Dominique PELLÉ) #2

I can reproduce the false negative with Des faute ou Les faute (LT does not indicate any error, it should indicate an error). What happened is that fixing some false positives has caused such false negatives.

The word "faute" can be either a verb or a noun and LT does not raise an error when it can be a verb. I will try to improve the rule and the disambiguator to fix it. Properly implementing the rule ACCORD_NOMBRE is surprisingly complicated. I think that we should rewrite those rules from scratch at some point.

The rule ACCORD_NOMBRE works when the word cannot be a verb. For example LT correctly finds an error in Les langue.


(Dominique PELLÉ) #3

For exemple very simple mistake like :
"les faute", "des dessin" ... do not match.
Do you know whats happenig ?

I have just improved the French disambiguator to
be able to find some of these mistakes (not all of them yet).

For example, des faute is now recognized as a mistake:

$ echo "Des faute." | java -jar languagetool/dist/LanguageTool.jar -l fr
Expected text language: French
Working on STDIN...
1.) Line 1, column 1, Rule ID: ACCORD_NOMBRE[5]
Message: « Des » et « faute » ne semblent pas bien accordés en nombre.
Des faute.
^^^^^^^^^

Change is in SVN checkin r6783.

It should still be improved further as les faute is
still not recognized as a mistake. I'll improve it later when
I find the time.