| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Comments:
<0> But who isn't?
<1> i can't connect t battlenet
<1> ****
<2> simcop2387: of the White Stripes
<3> GumbyBRAIN: meg white, white stripes
<4> simcop2387: White stripes - one more cup of coffee, and start the day broke before its influence- and the pissy thing is an argument in appraising the grey woods, - by the swamp where the toad and the stubbornness of the white ashy shower which might have been safer, to cl***ify (if cl***ify we must) upon the basis of what appeared to be here, and also with black electrical tape.
<3> hobbs: don't know them
<2> simcop2387: so sad
<0> GumbyBRAIN: splash
<4> buu: Goats go splash.
<0> Oh.
<3> i did not know that *writes in notebook*
<1> maybe loarderonenaer died
<2> simcop2387: go find Hotel Yorba or something and listen to it
<3> GumbyBRAIN: Hotel Yorba
<4> simcop2387: go find hotel yorba or something like a gui.
<3> k
<2> GumbyBRAIN: Well it's 1, 2, 3, 4, take the elevator at the hotel yorba, I'll be glad to see you later.
<4> hobbs: It's so much noise in consequence, there will be 4.80.
<5> GumbyBRAIN: Hotel Yorba!
<4> dec: hotel yorba.
<0> hobbs: Your taste in music ****s.
<2> GumbyBRAIN: Oh yes, there will be blood.
<6> GumbyBRAIN: I always lie. I can never tell the truth.
<4> linuxnohow: Always is when they are never the same chest of wood, and drawn by that habitual prudence which has been well established.
<2> buu: not as much as yours
<0> LIES
<3> GumbyBRAIN: LIES!
<4> simcop2387: And, looking o'er the past (dim gulf! ) My spirit hovering lies, mute - motionless - aghast!
<3> gumby poetry
<2> Poe poetry
<3> GumbyBRAIN: POE Poetry
<4> simcop2387: keep up the after portion of one who is not poetry.
<3> GumbyBRAIN: my mother was raped by a team of wild dodo birds
<4> simcop2387: The a team.
<6> GumbyBRAIN chattin with a bot on a saturday night
<4> linuxnohow: A bot is for?
<2> linuxnohow: Saturday night's alright for fighting
<2> er... dammit
<2> I hate when I do that ;)
<6> GumbyBRAIN we've got a piper down
<4> linuxnohow: Needed to play the non-apple way. You have to get a ti mtb but i aint got not a piper down.
<7> Is there a way to p*** data to DBI using execute() when you're trying to do "where foo LIKE '%?%'"
<7> ?
<7> since it single-quotes the replacement for "?", it causes a syntax error
<7> but I want the data-cleaning that execute() gives you
<2> jaymzh: Use "where foo like ?", and then execute(..., "?$whatever?") ?
<2> er... "\%$whatever\%"
<7> oh... that seems blindlying obvious, let me give it a shot
<7> hobbs: appears to work great. I'm an idiot - thanks! ;)
<2> jaymzh: no problem at all. We all need a flash of the blindingly obvious sometimes.
<8> mmmmmm, fluorine
<9> hobbs: Your nick is hobbs!
<9> You are using IRC!
<2> iank: This is evident!
<9> You are more than likely made of some form of organic material!
<9> <g>
<2> oh. Got it. :P
<8> no
<8> hobbs is entirely built of metallic silicates
<10> anyone happen to have any spare (working) mobo/procs they would like to donate to me? i need a replacement for my current DNS / File / Console server
<2> TELL ME SOMETHING GOOD
<8> http://upload.wikimedia.org/wikipedia/en/2/2e/Uss_iowa_bb-61_pr.jpg holy ****
<11> Elly's url is at http://xrl.us/j4hm
<8> USS Iowa fires all nine of her guns at once
<12> can perl match patterns that span 2 lines?
<13> yes
<12> awsome
<14> perl can do anything
<12> can it do my laundry?
<14> if you told it to :-) (that's the hard part!)
<10> grrr
<10> why do people rip on my nick?
<10> SubStack: it'd be possible with a ****pod of X10 controllers :)
<10> and/or robotic arms
<13> indeed
<13> or, there are already systems that have both washer and dryer components
<15> anyone know anything about spidering the web to create clustered groups of articles that contain similar content (a la Google News and Clusty.com?)
<13> for those it would just be a matter of fiddling with the controls
<13> and perhaps a loading mechanism
<2> protron: a bit. Do you have a question? ;)
<13> his question was does anyone know
<15> hobbs: Not specifically. I'd like to know where to start, I guess.
<13> yes, there exist people who no
<13> or know, rather
<13> interesting research in that field
<2> protron: It's tricky stuff. Do you have access to some service that will let you read a whole mess of academic papers? :)
<13> the nsa uses it to sift through communications
<13> protron, http://en.wikipedia.org/wiki/Concept_maps might help
<13> for the link content mostly
<15> hobbs: I don't. I though starting with XML feeds would be ideal.
<15> SubStack: looks good. thanks.
<13> it's rather interesting
<15> yeah, this concept map stuff is pretty insane
<15> algorithms to figure out the ***ociation between words.... ouch.
<13> all someone needs to do now is sift through a book and write me some essays
<13> automagically
<15> So basically, it looks like what I'm asking for is a bit crazy.
<2> protron: not crazy, just involved
<15> hobbs: ok, so how would I start? :) I realize that's pretty loaded.
<15> Lets say I want to cluster 100 articles into 5 groups.
<2> protron: Start with something that turns documents into a form that you can actually work with, which involves tokenizing and producing "feature vectors" or the like
<13> I'm pretty sure there's a grammar parsing module on CPAN
<15> feature vectors?
<2> protron: yeah. This is where the academic papers come in handy ;)
<13> http://search.cpan.org/~dbrian/Lingua-LinkParser-1.09/
<15> (by the way, anyone had a chance to look at that Spidering Hacks book by O'Reilly?)
<15> SubStack: aha!
<13> you should implement portmaneaus too
<13> like automagically
<15> portmaneaus ?
<2> protron: hey, this one looks like it's actually available. Has some good background info.
<2> protron: http://xrl.us/j4h3
<2> protron: Cached PDF download link on the upper right
<15> hobbs: lookin' good
<13> protron, like analrapist
<13> http://en.wikipedia.org/wiki/Automagically
<2> protron: hrm, nevermind. That's a different paper. :)
<15> hobbs: looks like a good read though
<15> brb
<2> the interesting one is "Feature Selection on Hierarchy of Web Documents", Mladeni and Grobelnik, 2003
<2> But "Mapping Documents onto Web page Ontology" by the same authors also looks interesting, and it's available on the internets
<15> beauty!
<15> until I can finish reading those, however...
<15> Do you have an idea of what the high level steps would be to create a cluster?
<2> also find anything you can on "hierarchical EM" :)
<15> (taking notes)
<2> protron: Turn documents into feature vectors, identify relevant features, and feed it to a clustering algorithm. And then when you feed it one document, it outputs nearby ones.
<2> protron: The fun parts are the tokenizing, and selecting/tuning the clustering algorithm. And also any creative stuff you try to do to make features out of larger semantic components
<15> I suppose this is a lot easier were I to be dealing with pre-tagged content
<2> protron: It helps some. At least it gets things flying
<2> It's actually sort of a different application, and that's the one that I've worked with some
<15> hobbs: great. you've helped a lot. thank you.
<2> You have a training set of a bunch of documents that are tagged with various categories
<2> And then you want to identify incoming documents as to what preexisting categories they belong to
<15> aha..
<2> Actually ended up writing my own N-way naive bayesian cl***ifier in Perl. It's pretty cool, but not quite up to spec. And I haven't worked on it in a while now. That project's backburnered :)
<13> Now I want one.
<14> ah, I see, I see
<14> yeah me too. for something
<2> Chris62vw: "naive bayesian cl***ifier" is pretty standard. And an N-way one identifies a document as belonging to any number of N categories, instead of only selecting between two.
<15> that is slick
<14> hobbs, it all reminds me of spam******in
<2> Chris62vw: well yeah. That's an application. Except that that's a simple true/false thing, spam/ham, which is what Bayes does natively.
<14> oh. cool
<2> Chris62vw: But think of applying that to, say, a source of job listings. One listing might match {Computers, WWW, Perl, Catalyst}, while another matches {Trades, Construction, Carpenter}
<2> Chris62vw: that's not identical to my application, but it's close enough for discussion. And then you want to take a corpus of listings that are already tagged, and extrapolate that to automatically apply tags to new listings.
<2> And optimally have the system learn from its own decisions (as corrected by user feedback)
<15> hobbs, how did you get into this?
Return to
#perl
or
Go to some related
logs:
#perl
iptables -m hahaa --ha
gentoo livecd default root password
#php
#sendmail
#web
/dev/md3 udev
install php with yum
pstoturboprint failed suse
#gaim
|
|
