[Libreoffice-bugs] [Bug 90006] spellcheck to indicate a spelling as rare thus usage to be confirmed

[bugzilla-daemon at bugs.documentfoundation.org]

https://bugs.documentfoundation.org/show_bug.cgi?id=90006

--- Comment #8 from Nick Levinson <Nick_Levinson at yahoo.com> ---
Thank you, both.

@Heiko Tietze:

No, but I think I see the problem. My use case is about rarities across general
language, not rarities that become common within one specialty, such as a
project. Consider that unabridged dictionaries (like Oxford Eng. Dict.) have
more entries than a spell checker has, like floccinaucinihilipilification,
because if someone types a rare word they're probably typing it by mistake and
it's a misspelling in context, and users want misspellings flagged, so spell
checkers usually omit rare words so they'll get flagged as misspellings. The
specialty dictionaries, once a word is added, would always approve a spelling,
so they're no help. Instead, what I'm proposing is for LO to notice that the
spelling is rare in general writing and therefore unlikely because normally it
would be an error but maybe it's not this time, so it asks the user. Maybe the
given spelling is correct only one time out of forty times that it's used but
is wrong all the other 39 times. That word in someone's essay should not be
marked as wrong or left unmarked as right but instead should be announced as
doubtful. (Example: A legal writer gave an essay to an assistant, who corrected
"justiciable" to "justifiable" when the original was right and the correction
was wrong. The original is rarely used even in law, which is why the assistant
didn't recognize it.) This would help a secretary who is supposed to fix their
boss's mistakes and would run mass replacement but should be stopped by LO when
a rarity is encountered, with LO treating the encountered string as rare and
questionable, not as right or wrong.

If Abble is a computer term or name apart from Apple, I didn't find it in
Google. If it's a misspelling, I'm not talking about pure misspellings.

@V Stuart Foote:

For Hunspell or Lightproof, this feature was not listed
(<http://hunspell.github.io/>,
<https://extensions.openoffice.org/en/project/lightproof-grammar-checker-development-framework>,
or https://extensions.openoffice.org/project/lightproof). If grammar means
syntax, Lightproof wouldn't be the vehicle for this feature.

I tested https://languagetool.org/ with floccinaucinihilipilification (with the
spelling confirmed in Google and which is in the Oxford English Dictionary (1st
ed.)) but the word was simply marked as wrong whether American or British was
selected. I also tested the sentence "I practice floccinaucinihilipilification
once in a while." The result was the same. That's not what I propose, because
"floccinaucinihilipilification" is a real but rare word and is not misspelled.
If LanguageTool had the feature I propose, it would have asked me about the
rarity in context. It didn't, so it doesn't.

I couldn't try https://www.grammarly.com/ Native Apps option, which is only for
Win or Apple (I have only Linux and *BSD), and, from what it says about the
free version, this feature is not included, although it might be in the
paid-for version.

The Boolean field I proposed was not "a non-maintainable data structure of
alternates". Entries can be maintained by users and others. Alternatives, if
they're what you meant, do not have to be provided; if the user believes that a
flagged word is an error in context, the user would replace it with whatever
word they like, but this feature need not suggest alternative words.

I wasn't thinking of a GUI (a hook would be needed for a CLI, too), but was
assuming that a field would be a hook that an extension couldn't add by itself.
If I'm wrong about that, so be it.

You seem to know the internals better than I do, so I won't change the status.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/libreoffice-bugs/attachments/20170728/7ff534b9/attachment.html>