Throw out low-quality questions

[Jorge Castro]

While it's nice to have the entire content of questions available, I think in the scope of a help lens we should try to cut out some junk on behalf of the user.

While using the lens I notice that it returns everything on the site, while that can be useful, it also clutters the lens with less-than-ideal results. Some ideas:

- Just toss out any question with a 1 or a 0. The community selects the best content anyway, by setting a baseline of 2 (Which isn't hard for quality content), we automatically drop the junk.
- Strongly favor highly voted results?

[Mark Tully (markjtully)]

The API provides both the number of upvotes and number of downvotes a question has (not answers, though, as far as I know). Questions marked as closed are already filtered out, so it is certainly possible to filter out questions with a minimum score as well. The results could then be arranged by number of votes.

Also, are there any tags that should be excluded from searches? Are questions tagged as lubuntu or xubuntu, for example, going to be useful to someone who is clearly using ubuntu (and unity), or do we run the risk of excluding too many useful answers by removing them?

[Jorge Castro (jorge)]

I think it'd exclude one important use case for tech support, where you have a friend using xubuntu/lubuntu asking you for help and you look it up on their behalf, or perhaps someone is interested in searching how another desktop does something.

I gave some thought as to locking the results to the version of ubuntu with the "12.04" tag and so on, as well as other tricks, but then I was thinking we'd overcomplicate it and end up with something worse... the site and community are already doing the quality filtering, the trick is just to figure out where we want that line to be.

BTW, I came to the number "2" as a baseline because it would require at least 2 people in the community upvoting to judge it as a quality question. I figure it's a decent place to start and then we can adjust later if need be. Would it make sense to have the minimum threshold as a gsettings value so we can have people try different scores and combinations?

[Mark Tully (markjtully)]

I've attached .deb containing a version of the askubuntu scope set to filter out results with less than 2 votes, for testing. After installing, kill the unity-scope-askubuntu process (which will be from the original version of the scope) The new version will then start automatically when the lens is activated.

As you say, there is a risk of ending up excluding useful results by adding too many restrictions to the searches that are run. Is there any statistical information available about Askubuntu that could help find a decent threshold for results? For example, number of questions (and %) per no of votes, or votes vs views. These could be useful to see how many questions would be excluded, and how popular they are.

A similar approach could be used with the asklibreoffice scope, since it has a similar voting mechanism. Although it has far less votes cast than on askubuntu.

[Jorge Castro (jorge)]

CCing James and Marco to see if this statistical information is available.

[James Gifford (jamesgifford)]

I haven't delved into the API as much as some, but I don't think there is a
way to do what you're looking for exactly, and I don't remember seeing an
API call for filtering by views... let me poke around and see what I can
find.

Cheers,
James Gifford
On Sep 12, 2012 8:45 PM, "Jorge O. Castro" <email address hidden> wrote:

> CCing James and Marco to see if this statistical information is
> available.
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1049376
>
> Title:
> Throw out low-quality questions
>
> Status in A Unity Lens for AskUbuntu.com:
> New
>
> Bug description:
> While it's nice to have the entire content of questions available, I
> think in the scope of a help lens we should try to cut out some junk
> on behalf of the user.
>
> While using the lens I notice that it returns everything on the site,
> while that can be useful, it also clutters the lens with less-than-
> ideal results. Some ideas:
>
> - Just toss out any question with a 1 or a 0. The community selects the
> best content anyway, by setting a baseline of 2 (Which isn't hard for
> quality content), we automatically drop the junk.
> - Strongly favor highly voted results?
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/askubuntu-lens/+bug/1049376/+subscriptions
>

[Mark Tully (markjtully)]

I've done a quick count of votes on askubuntu.com. Currently, there are 65644 questions in total. They break down as follows:

41461 have 1 or more votes (~63%)
26013 have 2 or more votes(~40%)
16142 have 3 or more votes (~25%)
10644 have 4 or more votes (~16%)
7430 have 5 or more votes (~11%)

This includes questions marked as [closed] on the site, by the way.