How about logging the failed "Go" queries on the server and ranking them? Searching the Wikipedia can be pretty frustrating at times, especially when you don't know exactly what you're looking for. You know what I mean, you *know* what you're looking for, you just don't know what exactly it is called, especially when that's an article which can have a lot of names.
Once the queries are logged and ranked, a relatively simple mechanism could be built to pick one of the top 100 every couple of hours and show it somewhere in every page, obviously as long as the article hasn't been already created. That way we increase the chances towards meaningful pathfinding (#REDIRECTs) and *really* "wanted articles".
I'm aware this would incur a tad more load on the server, but since people typically tend to follow links rather than search for stuff, I don't think this would be a serious performance hit. What do you think?
--Gutza
Gutza wrote:
How about logging the failed "Go" queries on the server and ranking them? Searching the Wikipedia can be pretty frustrating at times, especially when you don't know exactly what you're looking for. You know what I mean, you *know* what you're looking for, you just don't know what exactly it is called, especially when that's an article which can have a lot of names.
Once the queries are logged and ranked, a relatively simple mechanism could be built to pick one of the top 100 every couple of hours and show it somewhere in every page, obviously as long as the article hasn't been already created. That way we increase the chances towards meaningful pathfinding (#REDIRECTs) and *really* "wanted articles".
I'm aware this would incur a tad more load on the server, but since people typically tend to follow links rather than search for stuff, I don't think this would be a serious performance hit. What do you think?
A little extra table with a string and a counter (to count multiple ones), and maybe a timestamp, so we can automatically get rid of queries that were searched only once for in, say, a month. Or a week.
You probably don't mean to display /all top 100 queries/ somewhere on every page? ;-) IMHO a link to "Special:Most searched topics" would suffice...
As much as I like adding new toys (Phase II was practically drowning in 'em ;-) I'd suggest to put that aside until we have a monster server running the database again. Otherwise, I like it (had something similar, but more simple implemented once).
Magnus
Magnus Manske wrote:
Gutza wrote:
Once the queries are logged and ranked, a relatively simple mechanism could be built to pick one of the top 100 every couple of hours and show it somewhere in every page
A little extra table with a string and a counter (to count multiple ones), and maybe a timestamp, so we can automatically get rid of queries that were searched only once for in, say, a month. Or a week.
You probably don't mean to display /all top 100 queries/ somewhere on every page? ;-) IMHO a link to "Special:Most searched topics" would suffice...
[snipped]
Magnus
When I say "pick one" I obviously don't mean "display all". Is there any possibility that my original message could have been interpreted otherwise? I think it's easy to interpret correctly what I said: pick one, and show one; show *one* on *every* page. You're somehow twisting it into pick *all*, show *all* -- which would obviously result in showing them all on some isolated page... I was going towards a proposal which would have made *one* failed query available for *all* users in order for them to have a chance at fixing it somehow (either by creating a redirect or by creating an article).
--Gutza
Gutza wrote:
Magnus Manske wrote:
Gutza wrote:
Once the queries are logged and ranked, a relatively simple mechanism could be built to pick one of the top 100 every couple of hours and show it somewhere in every page
A little extra table with a string and a counter (to count multiple ones), and maybe a timestamp, so we can automatically get rid of queries that were searched only once for in, say, a month. Or a week.
You probably don't mean to display /all top 100 queries/ somewhere on every page? ;-) IMHO a link to "Special:Most searched topics" would suffice...
[snipped]
Magnus
When I say "pick one" I obviously don't mean "display all". Is there any possibility that my original message could have been interpreted otherwise? I think it's easy to interpret correctly what I said: pick one, and show one; show *one* on *every* page. You're somehow twisting it into pick *all*, show *all* -- which would obviously result in showing them all on some isolated page... I was going towards a proposal which would have made *one* failed query available for *all* users in order for them to have a chance at fixing it somehow (either by creating a redirect or by creating an article).
--Gutza
Just for the record, re-read your original post and realized I completely misinterpreted it when I sent the reply above. Typical online conversation misinterpretation (you know the drill, missing facial cues, etc). In any case, I apologize if I came out as a paranoid schizoid (which I probably did :-)).
You're right, I don't mean 100 queries/anywhere, I meant /one/ failed query /everywhere/, and that one query would only be picked once every two hours. I honestly don't think that would be any perceptible performance hit.
--Gutza
On Tue, Jan 13, 2004 at 03:33:54PM +0200, Gutza wrote:
How about logging the failed "Go" queries on the server and ranking them? Searching the Wikipedia can be pretty frustrating at times, especially when you don't know exactly what you're looking for. You know what I mean, you *know* what you're looking for, you just don't know what exactly it is called, especially when that's an article which can have a lot of names.
Once the queries are logged and ranked, a relatively simple mechanism could be built to pick one of the top 100 every couple of hours and show it somewhere in every page, obviously as long as the article hasn't been already created. That way we increase the chances towards meaningful pathfinding (#REDIRECTs) and *really* "wanted articles".
You could ask the operators for the apache-logs and run a simple script over them to compare them with the database dump. Just an idea.
ciao, tom
wikipedia-l@lists.wikimedia.org