Beefy Boxes and Bandwidth Generously Provided by pair Networks Cowboy Neal with Hat
good chemistry is complicated,
and a little bit messy -LW
 
PerlMonks  

Re: help with AI::Categorizer

by randyk (Parson)
on Nov 10, 2007 at 21:15 UTC ( #650084=note: print w/ replies, xml ) Need Help??


in reply to help with AI::Categorizer

Here's an example which uses the CPAN subject categories for the training set, and then classifies modules according to which category they probably best fit into:

use strict; use warnings; require AI::Categorizer; require AI::Categorizer::Learner::NaiveBayes; require AI::Categorizer::Document; require AI::Categorizer::KnowledgeSet; require Lingua::StopWords; # set up features: # - give different weights to subjects and bodies # - use stop words my %features = (content_weights => {subject => 2, body => 1}, stopwords => Lingua::StopWords::getStopWords('en'), stemming => 'porter', ); # this is the raw data to train with, which associates # numerical categories with subjects and bodies my $chaps = { 6 => {subject => q{Data Type Utilities}, body => q{Date Time Math List Tree Algorithm Sort}, }, 10 => {subject => q{File Names Systems Locking}, body => q{Directory Dir Stat cwd}, }, 12 => {subject => q{Opt Arg Param Proc}, body => q{Option Argument Argv Config Getopt}, }, 14 => {subject => q{Security and Encryption}, body => q{Authentication Crypt Digest PGP Des}, }, 15 => {subject => q{World Wide Web HTML HTTP CGI}, body => q{WWW Apache MIME Kwiki URI URL}, }, 17 => {subject => q{Archiving and Compression}, body => q{tar gzip gz zip bzip}, }, 18 => {subject => q{Images Pixmaps Bitmaps}, body => q{Chart Graphic}, }, 19 => {subject => q{Mail and Usenet News}, body => q{Sendmail NNTP SMTP IMAP POP3 MIME}, }, }; # create documents from $chaps to train with my $docs; foreach my $cat(keys %$chaps) { $docs->{$cat} = {categories => [$cat], content => {subject => $chaps->{$cat}->{subject}, body => $chaps->{$cat}->{body}, }, }; } my $c = AI::Categorizer->new( knowledge_set => AI::Categorizer::KnowledgeSet->new( name => 'CSL'), verbose => 1, ); while (my ($name, $data) = each %$docs) { $c->knowledge_set->make_document(name => $name, %$data, %features); } my $learner = $c->learner; $learner->train; # this is a test data set to categorize, # based on the training done above my $test_set = {'Math::Complex' => {content => {subject => q{Math}, body => q{Complex number data type} } }, 'Archive::Zip' => {content => {subject => q{Compression}, body => q{Interface to ZIP archive files} } }, 'Apache2::URI' => {content => {subject => q{Apache}, body => q{Perl API for manipulating URIs} } }, 'MIME::Lite' => {content => {subject => q{Mail}, body => q{Create MIME/SMTP mails w/attachements} } }, }; # see what category each element of $test_set gets put into, # using a threshold score of 0.9 my $threshold = 0.9; while (my ($name, $data) = each %$test_set) { my $doc = AI::Categorizer::Document->new(name => $name, content => $data->{content}, %features); my $r = $learner->categorize($doc); $r->threshold($threshold); my $b = $r->best_category; next unless $r->in_category($b); printf("%s is in category %d, with score %.3f\n", $name, $b, $r->scores($b)); }
This produces
Archive::Zip is in category 17, with score 0.998 Apache2::URI is in category 15, with score 0.917 MIME::Lite is in category 19, with score 1.000 Math::Complex is in category 6, with score 0.997


Comment on Re: help with AI::Categorizer
Select or Download Code
Re^2: help with AI::Categorizer
by glasswalk3r (Pilgrim) on Jan 04, 2013 at 14:14 UTC

    This is a quite old post, but maybe somebody could help with my doubt.

    I'm trying to use the threshold method as shown by randyk in the given example but the method is simply not working: doesn't matter what I give as a value, the method ignores the input. The AI::Categorizer::Hypothesis object $r does have a threshold attribute with a defined value, but how does it setup it is not clear in the documentation.

    Does anyone know how to define a threshold? I'm getting some results with lower scores that I don't want to work with.

    Alceu Rodrigues de Freitas Junior
    ---------------------------------
    "You have enemies? Good. That means you've stood up for something, sometime in your life." - Sir Winston Churchill

Log In?
Username:
Password:

What's my password?
Create A New User
Node Status?
node history
Node Type: note [id://650084]
help
Chatterbox?
and the web crawler heard nothing...

How do I use this? | Other CB clients
Other Users?
Others scrutinizing the Monastery: (19)
As of 2014-04-23 09:33 GMT
Sections?
Information?
Find Nodes?
Leftovers?
    Voting Booth?

    April first is:







    Results (541 votes), past polls