On Mon, Apr 22, 2013 at 9:54 AM, Cheng Xing <cxing561(a)gmail.com> wrote:
For example, if a programmer clicks the "Inspire
Me" button on Wikipedia,
articles such as the Whitespace programming language, Rubber Duck
Debugging, etc. would show up. Things that the user probably doesn't know
about, that would probably interest the user, will show up by clicking that
button. Very occasionally there'd be random things like Stitches, which
the user might know nothing about, but might actually be interesting.
Matt is right about how the Getting Started work relates to this. I think
it would be complementary rather than duplicative.
The big question for implementing a reader recommender system is where you
get your data from. Tracking reader activity on Wikipedia is a _very_
touchy subject for a whole host of legitimate reasons, and to be totally
honest I don't think you're going to be able to implement
any recommender system based on people's reading habits.
You might be able to do this if you limit yourself to recommending solely
based on the current page, or based on user input (e.g. a recommender
search with "What's your favorite article?" and some matching based on
various attributes you could tinker with.)
If you're interested in recommender systems, as Matt says we are nearly
ready to explore personalizing the list at Special:GettingStarted based on
your past editing history. This is a slightly smaller scope, so if you want
to take a whack at this, we might be able to help you along the way with
advice, code review, etc. (assuming the engineers on the team don't think
I'm crazy).
Steven