[Foundation-l] "RevisionRank": automatically finding out high-quality revisions of an article

henna henna at wikipedia.be
Tue Dec 20 14:33:33 UTC 2011


On Tue, Dec 20, 2011 at 14:55, Yao Ziyuan <yaoziyuan at gmail.com> wrote:
> On Tue, Dec 20, 2011 at 6:51 PM, Tobias
> <church.of.emacs.ml at googlemail.com>wrote:
>
>> On 12/19/2011 11:38 PM, Yao Ziyuan wrote:
>> > Hi Wikipedians,
>> >
>> > I seem to have found a way to automatically judge which revision of a
>> > Wikipedia article has the best quality.
>> >
>> > It's very simple: look at that article's edit history and find out,
>> within
>> > a specified time range (e.g. the past 6 months), which revision remained
>> > unchallenged for the longest time until the next revision occurred.
>>
>> Hey Ziyuan,
>>
>> that's great! Have you made a statistical analysis whether the average
>> revision that remained unchallanged by a long time is better than the
>> average other revisions?
>>
>
> Honestly I haven't done a lot of tests, but I did investigated how
> "ultra-stable Linux distributions" (Debian, RHEL/CentOS) select stable
> software packages. I found Debian's model very similar to my idea: latest
> software package versions are put in a pool called "unstable"; if a package
> version remains in the "unstable" pool for a certain period of time with no
> serious bugs discovered, it is automatically moved to the next pool,
> "testing"; again, if a package version remains in the "testing" pool for a
> certain period of time with no bugs discovered, it is automatically moved
> to the next pool, "stable". Each new major release of Debian is a
> collection of all packages in the "stable" pool. (
> http://en.wikipedia.org/wiki/Debian#Development_procedures)
>
> So it seems this "trial by time" approach at least works for a big open
> source software project like Debian.
>
>

It works because in order to put packages in the testing/unstable pool
you need to have 'Debian Developer' credentials. Getting those
credentials is harder then getting an account on Wikipedia and hence
the input into testing&stable can be considered a best effort and
doens't have to account for potential vandalism etc.



More information about the foundation-l mailing list