I just read about an open-source MySQL plugin that enables fulltext search on all MySQL table types:
On the other hand, our current search seems to hold up nicely these days...
Just FYI, Magnus
On 6/27/06, Magnus Manske magnus.manske@web.de wrote:
On the other hand, our current search seems to hold up nicely these days...
By what definition of hold up? Yes it's does seem to be providing results faster than ever but they haven't stopped being low quality and nearly useless.. The bug that where certain characters in the excerpt halt the display of more results still exists as well.
On 27/06/06, Gregory Maxwell gmaxwell@gmail.com wrote:
By what definition of hold up? Yes it's does seem to be providing results faster than ever but they haven't stopped being low quality and nearly useless..
You mean the indexing's too out of date, or something else?
The bug that where certain characters in the excerpt halt the display of more results still exists as well.
Is this known on BugZilla?
Rob Church
On 6/27/06, Rob Church robchur@gmail.com wrote:
On 27/06/06, Gregory Maxwell gmaxwell@gmail.com wrote:
By what definition of hold up? Yes it's does seem to be providing results faster than ever but they haven't stopped being low quality and nearly useless..
You mean the indexing's too out of date, or something else?
The last index update seems to be somewhere between 10th and 20th May. Could be better.
Meanwhile, could I request that redirects be deprioritised in search results? For example, search for "Couesnon". The first two matches are redirects to "Couesnon River", the actual article, which appears 3rd.
Steve
On 27/06/06, Steve Bennett stevage@gmail.com wrote:
On 6/27/06, Rob Church robchur@gmail.com wrote:
On 27/06/06, Gregory Maxwell gmaxwell@gmail.com wrote:
By what definition of hold up? Yes it's does seem to be providing results faster than ever but they haven't stopped being low quality and nearly useless..
You mean the indexing's too out of date, or something else?
The last index update seems to be somewhere between 10th and 20th May. Could be better.
We're aware of that problem. Brion's been working on improvements to the dump process; once those are done, I suspect automation will be adopted. When this happens, more frequent updates will be possible.
I've also quizzed him of late, and I am of the impression that better indexing is on the agenda too.
Meanwhile, could I request that redirects be deprioritised in search results? For example, search for "Couesnon". The first two matches are redirects to "Couesnon River", the actual article, which appears 3rd.
That's possible.
Rob Church
On 6/27/06, Rob Church robchur@gmail.com wrote:
Meanwhile, could I request that redirects be deprioritised in search results? For example, search for "Couesnon". The first two matches are redirects to "Couesnon River", the actual article, which appears 3rd.
That's possible.
Is it worth my bugzilla'ing?
Steve
On 27/06/06, Steve Bennett stevage@gmail.com wrote:
On 6/27/06, Rob Church robchur@gmail.com wrote:
Meanwhile, could I request that redirects be deprioritised in search results? For example, search for "Couesnon". The first two matches are redirects to "Couesnon River", the actual article, which appears 3rd.
That's possible.
Is it worth my bugzilla'ing?
I think it's been done. :)
Rob Church
On 6/27/06, Rob Church robchur@gmail.com wrote:
On 27/06/06, Gregory Maxwell gmaxwell@gmail.com wrote:
By what definition of hold up? Yes it's does seem to be providing results faster than ever but they haven't stopped being low quality and nearly useless..
You mean the indexing's too out of date, or something else?
Indexing is out of date, searching for multiple words gives the union not the intersection, if there is stemming it is broken, and no ability to do fuzzy matching (no soundex or double metaphone equal) means that if you don't spell something the same as the article (or you don't know how to spell it, a top use for search...) then you don't get results.. oh and searching other namespaces is broken ( for example http://en.wikipedia.org/wiki/Special:Search?ns8=1&search=There&fullt...).
Due to these factors it is always more useful to use google to search Wikipedia.
The bug that where certain characters in the excerpt halt the display of more results still exists as well.
Is this known on BugZilla?
I'm not sure. It's easy to reproduce, but hard to pin down when things aren't up to date. I recall that Kate knew about it. ::shrugs::.
On 6/27/06, Gregory Maxwell gmaxwell@gmail.com wrote:
Due to these factors it is always more useful to use google to search Wikipedia.
Out of curiosity, why do the links to search Wikipedia via google and MSN only appear when the internal search function is down?
Steve
Steve Bennett wrote:
On 6/27/06, Gregory Maxwell gmaxwell@gmail.com wrote:
Due to these factors it is always more useful to use google to search Wikipedia.
Out of curiosity, why do the links to search Wikipedia via google and MSN only appear when the internal search function is down?
Because their purpose is as a fallback when search is down.
-- brion vibber (brion @ pobox.com)
On Tue, Jun 27, 2006 at 05:13:15PM -0400, Gregory Maxwell wrote:
searching for multiple words gives the union
not the intersection,
I wouldn't want anyone to think I'm going off on a rant here... (;-) but if the default for multi word searches *is* OR instead of the generally expected AND, someone please confirm that for me so I can go file a bug.
Cheers, -- jra
On 6/27/06, Jay R. Ashworth jra@baylink.com wrote:
I wouldn't want anyone to think I'm going off on a rant here... (;-) but if the default for multi word searches *is* OR instead of the generally expected AND, someone please confirm that for me so I can go file a bug.
"Hippopotamus" turns up search results, "cephalopod" turns up search results, "hippopotamus cephalopod" turns up no results. The search appears to be intersection, not union, as expected. When I tried "hippopotamus cat", however, some of the excerpts didn't include both words; I had to go to the article and search for the missing word to verify that both were indeed present.
On 6/27/06, Simetrical Simetrical+wikitech@gmail.com wrote:
On 6/27/06, Jay R. Ashworth jra@baylink.com wrote:
I wouldn't want anyone to think I'm going off on a rant here... (;-) but if the default for multi word searches *is* OR instead of the generally expected AND, someone please confirm that for me so I can go file a bug.
"Hippopotamus" turns up search results, "cephalopod" turns up search results, "hippopotamus cephalopod" turns up no results. The search appears to be intersection, not union, as expected. When I tried "hippopotamus cat", however, some of the excerpts didn't include both words; I had to go to the article and search for the missing word to verify that both were indeed present.
I tested before making the claim in the first place, and the first article that came up didn't include all the words... but it used to. I'll go troll elsewhere. :(
On 6/27/06, Jay R. Ashworth jra@baylink.com wrote:
On Tue, Jun 27, 2006 at 05:13:15PM -0400, Gregory Maxwell wrote:
searching for multiple words gives the union
not the intersection,
I wouldn't want anyone to think I'm going off on a rant here... (;-) but if the default for multi word searches *is* OR instead of the generally expected AND, someone please confirm that for me so I can go file a bug.
It looks like I was wrong there, my apologies.
On Tue, Jun 27, 2006 at 10:13:07PM -0400, Gregory Maxwell wrote:
On 6/27/06, Jay R. Ashworth jra@baylink.com wrote:
On Tue, Jun 27, 2006 at 05:13:15PM -0400, Gregory Maxwell wrote:
searching for multiple words gives the union
not the intersection,
I wouldn't want anyone to think I'm going off on a rant here... (;-) but if the default for multi word searches *is* OR instead of the generally expected AND, someone please confirm that for me so I can go file a bug.
It looks like I was wrong there, my apologies.
Based on the description I should have been sensible enough to go produce myself before shooting my mouth off, it sounds like it's the AND, but applied to *entire articles*, not just article titles, which is a) perfectly sensible, and b) could be interpreted to look like something else, if you weren't thinking clearly.
Which I'm not; no A/C in the house, nor an exhaust fan. Florida. June.
Cheers, -- jra
wikitech-l@lists.wikimedia.org