\[~ann.campbell.2], as discussed with [~nicolas.peru] today, this rule generates some noise/controverses out of the box and even with **.api.** as default value for parameter forClasses. So we would suggest to remove it from the "Sonar way" quality profile.
\[~yuhan0501] That particular section of the rule has been relaxed in SonarJava 4.13 (it doesn't check the number of words any longer) but we forgot to update the RSPEC itself, I apologise for that oversight. I now have updated the RSPEC to reflect that decision.
And you are right, previous implementation had a bug where it ignored localisation.