[Foundation-l] How to improve quality of Wikipedia?

Andrew Gray andrew.gray at dunelm.org.uk
Sun Oct 10 11:06:09 UTC 2010

2010/10/10 Виктория <mstislavl1 at gmail.com>:
> *Dzień dobry, *Przykuta
> One of Wikipedia perennial dilemmas is quantity vs. quality. Low depth and
> low articles to non-auricles ratio usually a sign that too many articles
> were created semiautomatically, by bots and the community is spread too thin
> e.g. there is not enough people to correct and discuss these articles.

Polish doesn't seem to have more bots *editing* than other projects
do. A few months back, I graphed all the Wikipedias by number of bot
edits as proportion of total edits:


Polish isn't marked here, but it's eighth from the right - it doesn't
seem to be a statistical outlier at all. Unless the bots are
concentrated solely on new articles, which is a possibility, this
seems normal.

So perhaps it's something about the way the Polish Wikipedia works? A
few thoughts:

* Polish doesn't host any images - unlike most other projects - so
there's no need for image pages, image talkpages, etc. On some
projects, such as German, as many as 6% of pages are in the image

* Polish doesn't seem to use article talkpages much. I've just spent
some time hitting "Losuj artykuł", and about 10-20% of the articles I
found had talkpages. In English, this is about 85-90%, and in French,
about the same. In the other languages these may just have project
tags ("this article is part of WikiProject Something") or metadata
("this article is rated C-class and needs an image"), but they still
show up as non-article pages. There's currently ~735,000 articles and
~595,000 non-articles; if another 70% of articles were to have
talkpages - making it comparable with English and French - this would
make ~1,110,000 non-articles, or 1.5 non-articles per article.

* Finally, Polish Wikipedia has fewer active users than any of the
next three "smaller" Wikipedias - Italian, Japanese and Spanish -
which might be significant here. Fewer users talk less, so there's
fewer "natural" discussion pages.

> You can also have an "X week" where X is any topic of articles created by
> bots. People like to work together on a common goal, in the Russian
> Wikipedia thematic weeks are very successful.

English Wikipedia has had some success with a "cup" system - a hundred
Wikipedians competing over several months to improve articles, etc.
It's hard to say how much impact it's had, or how much work people
would have done *without* the contest, but I've seen estimates that a
quarter or a third of all highly-rated content over the last year has
come from participants. In some cases, it was so popular it
overwhelmed the review processes!


I've not participated in theme weeks before, but I've heard pretty
good things about them. Were they usually focused on creating articles
or on "saving" existing ones?

> And lastly you can start nominating articles created by bots and not touched
> by a human hand since then for deletion. They will be either improved or
> deleted and any outcome will increase average depth. In RuWiki nobody tries
> to nominate significant bot articles like German cities but superfluous ones
> about obscure 70s C-movies and far far away galaxies NGO... are nominated
> for deletion 5 per day.

Harsh but fair!

How strict is the bot-approval process on Polish Wikipedia? If there's
a problem with mass creation of articles, you could try being stricter
about requiring community approval before the bots are allowed to run,
to check that you actually do want these topics.

- Andrew Gray
  andrew.gray at dunelm.org.uk

More information about the foundation-l mailing list