OK
AJAX error!

Les forumsDictionnaireImprovements for French spell checker in Hunspell

Improvements for French spell checker in Hunspell

I have been looking into the Hunspell support for French and I have found a few improvements.

a) single characters such as '_' '&' 'α' 'β' 'γ' do not need to be in the .dic file, see www.dicollecte.org…

b) Buenos Aires is included with an underscore, see www.dicollecte.org…

c) The single character word 'l' has been included two times, see cgit.freedesktop.org… and also the .dic file in the Debian package.

d) Both WORDCHARS and TRY can be extended with characters such as '²' '₂' 'ᵉ' '.' '-' (Note that in Nuspell, support for WORDCHARS might disappear, hence make TRY as complete as possible.)

Sources:
- cgit.freedesktop.org…
- cgit.freedesktop.org…
- packages.debian.org…

Please notify LibreOffice and Debian when a new version has been made.

For more information on Nuspell, Hunspell's successor with backwards compatibility, please follow:
- github.com…
- www.facebook.com…
- twitter.com…
le 14 juin 2018 à 14:31

pander :
a) single characters such as '_' '&' 'α' 'β' 'γ' do not need to be in the .dic file, see www.dicollecte.org…


Why not? They are useful at the moment.
If I type some unknown chars from the dictionary, like ŵ, they are underlined as it should.
Actually, some entries in the French Hunspell dictionary are mostly useful for Grammalecte, whose lexicon is converted from the Hunspell dictionary.

b) Buenos Aires is included with an underscore, see www.dicollecte.org…


Again, that’s not useful for Hunspell, but it is for Grammalecte. That’s a trick. But you’re right, it’s useless for the Hunspell dictionary. I should remove it.

c) The single character word 'l' has been included two times, see cgit.freedesktop.org… and also the .dic file in the Debian package.


Yes, but with different tags for the grammar checker. Until now, it has never been a problem for Hunspell. Is it now?
For years, the Hunspell dictionary was provided with grammatical tags, and there was several similar words with different tags. Now, the Hunspell dictionary is provided without these tags, for the grammar checker now uses its own spellchecker (written in pure Python or JavaScript).

d) Both WORDCHARS and TRY can be extended with characters such as '²' '₂' 'ᵉ' '.' '-' (Note that in Nuspell, support for WORDCHARS might disappear, hence make TRY as complete as possible.)


OK.

Please notify LibreOffice and Debian when a new version has been made.


I will for LibreOffice. But for Debian, what’s the point? AFAIK, they usually copy/paste dictionaries from the LibreOffice repo.

For more information on Nuspell, Hunspell's successor with backwards compatibility, please follow:
- github.com…
- www.facebook.com…
- twitter.com…


Thanks.
Will Nuspell be bundled within LibreOffice, Firefox, Thunderbird, Linux…?
Will Hunspell be deprecated when Nuspell is ready?
le 15 juin 2018 à 14:47

</p>

Please notify LibreOffice and Debian when a new version has been made.


I will for LibreOffice. But for Debian, what’s the point? AFAIK, they usually copy/paste dictionaries from the LibreOffice repo. <p>



Actually the Debian package is taken directly from dicollecte.org, not from LibreOffice.

However no need to ping us, we have a script that automatically detects new versions of the dictionary when they are released.

Best,
le 17 juin 2018 à 09:23
Thanks for the information.
le 17 juin 2018 à 11:48
Thanks for the feedback. Nuspell is being developed independently from Hunspell from now on. What will happen to Hunspell is up to its maintainer. Nuspell will support an identical API, so it can be used as a drop-in replacement. It is still in development and once it reaches beta we will contact people that make ports, packages and integrations. For following us on GitHub is a new link: github.com…
le 07 juillet 2018 à 18:18

Notification par e-mail    0