Topic on Talk:Wikimedia Apps/Team/Android/AppEditorTasks

Hi,

I see a lot of bad description added with Suggestededit. Especially in typography, description in French starting with an uppercase letter (which should rarely happens) or ending with a punctuation (which should almost never happened). Couldn't these mistakes be removed?

Hi, thanks. We're working on quality controls and aware they're not as good as we wanted them to be. I'll bring up this specific issue.

Hi,

Thanks, it would be both great and easy to fix this problem.

I agree: I often see descriptions wrongly starting with an uppercase letter or ending with punctuation. Is there any update about these two problems? Thanks!

+1.

The suggestion are not really bad (I changed the title), the content itself is globally good enough, it's only the typography (and only of the first and last character).

Here the solution seems pretty simple and could be something like:

• if the suggestion ends with a punctuation (or start with an uppercase), remove the punctuation, except for a whitelist of the few exception where the punctuation is ok (there is almost no exception for punctuation, maybe a lot and tricky ones for uppercase...).

Perfectly agree

https://phabricator.wikimedia.org/T251114

Capitalisation sounds really tricky. For example:

"Poet and playwright from Quebec"

"Quebecois poet and playwright"

One of these shouldn't be capitalised, one of them can take the capital letter. And then it would vary from language to language – unless I'm mistaken, neither of these should capitalised in French, nor should they in Swedish, but that just makes it more difficult since we have so many languages.

Yes it can tricky. There is even more tricky cases like "church of England" (for a place) and "church of England" (for an organisation).

That said, in most languages and in the majority of cases, it's ore probable to start with a lowercase than with an uppercase. If we go for a simple rule: "all in lowercase" is better than "all in uppercase" but of course there is a lot of exception (German is almost always uppercase for instance).

A better but more complex rule could be something like "look at the value of P31 and if the first word is the same as the label of the value of P31 then follow this casing". This is basically the query I do everyday for correcting labels and description of churches and Churches ;)

My reply in https://phabricator.wikimedia.org/T255985 is probably relevant here too:

These are our current plans to address the specific problem of capital letters and punctuation in Wikidata descriptions:

• Adding another step to suggested edits onboarding in the app specifically about this
• A warning when the user starts the sentence with a capital letter – this might not be desired behaviour, sure you want to continue?
• A warning when the user ends the sentence with punctuation – this might not be desired behaviour, sure you want to continue?

We're also looking into long-term solutions for more general issues. One major thing we have yet to solve is re-working detection of bad edits (this is currently based on undoing and reverting of the editors edits, but that proved ineffective). We're also doing some other investigations to see how we can ease the burden of patrolling.