https://bugs.documentfoundation.org/show_bug.cgi?id=95274

--- Comment #55 from [email protected] ---
I second Stéphane's suggestions.

The current behaviour is weird at best. The word "line" suggests Friulian as
the language, "car" results in Romanian, "test" in Estonian, ...

Does libexttextcat score/rate its guesses? I'd expect things to go wrong with
single words. So, either discard language guesses with a low score, or guesses
for paragraphs with less than e.g. five words.




Although, even with complete sentences we are not guaranteed to get correct
suggestions. The following text is recognised as "Norwegian, Nynorsk", although
it is written in Bokmål: "Dette er ikke nynorsk. Du gjettet feil."

Translated into Nynorsk, it would be: "Dette er ikkje nynorsk. Du gjetta feil."

Notice how "ikke" and "gjettet" are exclusively Bokmål, while "ikkje" and
"gjetta" are Nynorsk.




It would be nice if LO would indicate that a language is a suggestion (from
libexttextcat). The language lists (in "Tools > Language > For
Selection/Paragraph/All Text") could be expanded to something like this:

=================================
Language(s) in current document:
English (UK)
☑ French (France)
More ...
---------------------------------
Suggested for current paragraph:
Norwegian (Bokmål)
---------------------------------
None (Do not check spelling)
---------------------------------
Reset to Default Language
=================================

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to