Log of the #blacklight channel on chat.freenode.net

Using timezone: GMT-05:00
* g8tor_ leaves00:40
* dcage joins09:35
* rsinger leaves10:16
* rsinger joins10:19
* dcage leaves10:23
<cbeer>BillDueber: that edismax thing was too easy.. guess i'll have to start on implementing designs now10:31
<BillDueber>cbeer: Glad it came together for you! I'm actually having some problems with it right now....dammit.
<cbeer>oh? what should i watch out for?10:33
<erikhatcher>BillDueber: what kinds of problems?10:36
<BillDueber>erikhatcher: Dude, you're *everywhere*! :-)
erikhatcher: I have a field type that I call "Exactmatcher" -- it takes out spaces, punctuation, lowercases everything, downgrades diacritics, etc. For index & query10:37
erikhatcher: But it doesn't seem to work as a pf field, which threw me off.
erikhatcher: Actually, contines to throw me off.
<erikhatcher>can you show a simple debugQuery output?10:38
<BillDueber>erikhatcher: If...you tell me how to do that. Then, yes!
* BillDueber has giant, gaping hole in his knowledge about pretty much everything.10:39
<erikhatcher>&debugQuery=true10:40
look at the response and see what it parses the query to
<BillDueber>erikhatcher: OK. I'm on it.
erikhatcher: OK, so I guess I had a fundamental misundersatnding about dismax10:52
I'm seeing from the debug output that the "phrase query" is really doing bigrams. Not the whole thing.
(this is with the edismax)
<erikhatcher>BillDueber: do you have a bigramming field setup? 10:53
what do you get if you switch to pure dismax?10:54
<BillDueber>erikhatcher: Not to the best of my knowledge. Certainly not on purpose.
erikhatcher: Lemme restart it and check.
<erikhatcher>BillDueber: you should be able to do &debugQuery&defType=dismax or defType=edismax to compare the two10:55
no restart needed to toggle query parsers10:56
<BillDueber>erikhatcher: With straight-up dismax, I'm *not* seeing the bigrams - -just the whole phrase, as expected.10:59
erikhatcher: I'm using the edismax patch in jira against 1.4 stable11:00
<erikhatcher>BillDueber: and you're using exactly the same config except for the defType parameter?
<BillDueber>erikhatcher: Yup.11:01
<erikhatcher>BillDueber: now it's example time... show me a parsed query from each
<BillDueber>erikhatcher: OK.
@help lisppaste
erikhatcher: http://paste.lisp.org/+1YNY11:12
<erikhatcher>BillDueber: ok, i'm looking into edismax now.... 11:16
BillDueber: looks like an issue you'll want to bring up on solr-user
<BillDueber>erikhatcher: Yeah. Dammit
<erikhatcher>BillDueber: are you removing stop words?
<BillDueber>erikhatcher: No
erikhatcher: Not in the data this one is hitting, anyway.
<erikhatcher>BillDueber: your results are the same amount.... i'm not sure what's not working right for you11:17
<BillDueber>erikhatcher: The ranking
erikhatcher: If someone searches for "gone with the wind", an item with the title "The wind" shouldn't appear on the first page.11:18
erikhatcher: It toally screws up my ability to boost complete title matches
<erikhatcher>BillDueber: yeah, i see what you mean11:19
BillDueber: best to post this very example to solr-user... yonik will surely respond
<BillDueber>erikhatcher: Writing now -- thanks!
erikhatcher++ yet again
<erikhatcher>nah, no increment for me on this one... i'm scratching my head!
<BillDueber>erikhatcher: Yeah, but do you have any idea how long I would have screwed with this trying to figure out what was going on without you? The answer is "too damn". Too damn long.11:20
<erikhatcher>query parsing - no trivial job to do "right"11:21
* tachyonwill_ joins
* jkeck joins12:35
<BillDueber>cbeer, erikhatcher: I patched to allow a pf1 arugment (turn query into a single phrase, not bigrams). Now it works as I'd expect. Wish I understood the code I was writing, though...13:40
<erikhatcher>BillDueber: you and me both, man13:42
BillDueber++
<ndushay>BillDueber: holy @$!#$ - you already fixed that problem????14:14
will your fix be avail?
I need to use that patch - our advanced search isn't recognizing phrase searches.
<BillDueber>ndushay: Well, part of the issue is that not everyone agrees it's a problem. I imagine if you're not working with something as complex as bib data, what they're doing is smart.
I'll build a patch against 1.414:15
<ndushay>i'm not sure I fully understand your special field
<BillDueber>ndushay: My confidence that I'm not doing something stupid is ....low.
<ndushay>no wait, i do understand
you're saying "if this is an EXACT match with the words in the exact-match field, then boost the heck out of it"
i got around that by14:16
having unstemmed and stemmed versions of fields
boosting the unstemmed higher
and really boosting in pf
not sure if that's better or not.
<BillDueber>ndushay: It should be the same, depending on how aggressive you are about eliminating punctuation and diacritics and such.14:17
<ndushay>right. we get rid of that stuff.
i can hardly wait to tackle CJK, though. NOT
<BillDueber>ndushay: Yeah, I'm not sure how much of the magic I'm doing is necessary with a move to dismax.
Need to think hard about it, and my head just isn't in the game today.14:18
<ndushay>BillDueber: let dismax magic work for you
<BillDueber>ndushay: Right. I just need to make sure I'm not losing anything. 14:19
<ndushay>is this vufind?
<BillDueber>ndushay: Only kindof.
:-)
<ndushay>is there a way to do relevancy testing, the way we use cucumber for our blacklight relevancy testing?
'cause that would be the best way to tell.
<BillDueber>ndushay: Sure. Get two people who agree on relevancy for a given search and get back to me.
<ndushay>i have tests like this:
"results should have these 3 resources in top 5 hits"14:20
I get the specifics from feedback emails - every time someone complains about a particular search
<BillDueber>ndushay: Except sometimes they're wrong.
<ndushay>and also when i am trying to figure out certain things and have to find my own searches.
sure - but I let OTHER librarians catch that
or I figure it out and we negotiate.
<BillDueber>ndushay: I've got one prof who keeps insisting that a search on 'mind' should have the journal show up first -- except that's not the full name of the jouranl, and there are plenty of items whose entire title is 'Mind'.14:21
<ndushay>ah
that's a query time boost fix.
<BillDueber>ndushay: But, of course, that's an outlier.
ndushay: And only if he's logged in :-)
<ndushay>right! "if jerkface is searching for X, then use a function(?) query to manipulate the results this way"
right to his IP address.14:22
heh.
a new standard for library discovery software.
actually, it would be kinda cool if you could adjust query time boost in your personal profile ...
<BillDueber>ndushay: Yeah. Esp. for subjects or our High Level Browse stuff -- general stuff like that.14:23
* ndushay leaves14:39
* ndushay joins
<cbeer>not sure if any of you guys are doing this, but with the solr DIH, is it possible to run an xsl transform using the FileDataSource? I see some stuff about xsl + urldatasource or the xpath processor, but.. 14:59
* g8tor leaves16:35
* tachyonwill_ leaves17:00
* BillDueber leaves17:02
* jkeck leaves17:56
* jkeck joins17:57
* jkeck leaves19:28
* cbeer_ joins19:31
* wlayton joins20:37
* cbeer_ leaves21:35
* ndushay leaves21:44
* rsinger leaves22:05
* wlayton leaves23:21

Generated by Sualtam