# WordNet::Similarity::lesk.pm version 2.04
# (Last updated $Id: lesk.pm,v 1.29 2008/03/27 06:21:17 sidz1979 Exp $)
#
# Module to accept two WordNet synsets and to return a floating point
# number that indicates how similar those two synsets are, using an
# adaptation of the Lesk method as outlined in <ACL/IJCAI/EMNLP paper,
# Satanjeev Banerjee, Ted Pedersen>
#
# Copyright (c) 2005,
#
# Ted Pedersen, University of Minnesota Duluth
# tpederse at d.umn.edu
#
# Satanjeev Banerjee, Carnegie Mellon University, Pittsburgh
# banerjee+ at cs.cmu.edu
#
# Siddharth Patwardhan, University of Utah, Salt Lake City
# sidd at cs.utah.edu
#
# This program is free software; you can redistribute it and/or
# modify it under the terms of the GNU General Public License
# as published by the Free Software Foundation; either version 2
# of the License, or (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to
#
# The Free Software Foundation, Inc.,
# 59 Temple Place - Suite 330,
# Boston, MA 02111-1307, USA.
#
# ------------------------------------------------------------------
package WordNet::Similarity::lesk;
=head1 NAME
WordNet::Similarity::lesk - Perl module for computing semantic relatedness
of word senses using gloss overlaps as described by Banerjee and Pedersen
(2002) -- a method that adapts the Lesk approach to WordNet.
=head1 SYNOPSIS
use WordNet::Similarity::lesk;
use WordNet::QueryData;
my $wn = WordNet::QueryData->new();
my $lesk = WordNet::Similarity::lesk->new($wn);
my $value = $lesk->getRelatedness("car#n#1", "bus#n#2");
($error, $errorString) = $lesk->getError();
die "$errorString\n" if($error);
print "car (sense 1) <-> bus (sense 2) = $value\n";
=head1 DESCRIPTION
Lesk (1985) proposed that the relatedness of two words is proportional to
to the extent of overlaps of their dictionary definitions. Banerjee and
Pedersen (2002) extended this notion to use WordNet as the dictionary
for the word definitions. This notion was further extended to use the rich
network of relationships between concepts present is WordNet. This adapted
lesk measure has been implemented in this module.
=head2 Methods
=over
=cut
use strict;
use warnings;
use Text::OverlapFinder;
use WordNet::Similarity;
use File::Spec;
use WordNet::Similarity::GlossFinder;
our @ISA = qw(WordNet::Similarity::GlossFinder);
our $VERSION = '2.04';
WordNet::Similarity::addConfigOption ("normalize", 0, "i", 0);
=item $measure->initialize($file)
Overrides the initialize method in the parent class (GlossFinder.pm). This method
essentially initializes the measure for use.
Parameters: $file -- configuration file.
Returns: none.
=cut
# Initialization of the WordNet::Similarity::lesk object... parses the config file and sets up
# global variables, or sets them to default values.
# INPUT PARAMS : $paramFile .. File containing the module specific params.
# RETURN VALUES : (none)
sub initialize
{
my $self = shift;
my $paramFile;
my $wn = $self->{wn};
my $class = ref $self || $self;
# Stemming? Normalizing?
$self->{stem} = 0;
$self->{normalize} = 0;
# Look for the default lesk relation file...
if(!defined $self->{relationDefault})
{
my $path;
my $header;
my @possiblePaths = ();
# Look for all possible default data files installed.
foreach $path (@INC)
{
# JM 1-16-04 -- modified to use File::Spec
my $file = File::Spec->catfile($path, 'WordNet', 'lesk-relation.dat');
push @possiblePaths, $file if(-e $file);
}
# If there are multiple possibilities, get the one in the correct format.
foreach $path (@possiblePaths)
{
next if(!open(RELATIONS, $path));
$header = <RELATIONS>;
$header =~ s/\s+//g;
if($header =~ /RelationFile/)
{
$self->{relationDefault} = $path;
close(RELATIONS);
last;
}
close(RELATIONS);
}
}
# Call the initialize method in the parent...
$self->SUPER::initialize (@_);
# initialize string compare module. No stemming in string
# comparison, so put 0.
#&string_compare_initialize(0, %stopHash);
my @finder_args = ();
if (defined $self->{stop}) {
push @finder_args, stoplist => $self->{stop};
}
# lesk doesn't use a comp file, so we can ignore that
$self->{finder} = Text::OverlapFinder->new (@finder_args);
}
=item $lesk->traceOptions()
This method is internally called to determine the extra options
specified by this measure (apart from the default options specified
in the WordNet::Similarity base class).
Parameters: none.
Returns: none.
=cut
# 12/5/03 JM (#1)
# show all config options specific to this module
sub traceOptions
{
my $self = shift;
$self->{traceString} .= "normalize :: $self->{normalize}\n";
$self->SUPER::traceOptions();
}
=item $lesk->getRelatedness
Computes the relatedness of two word senses using the Extended Gloss
Overlaps algorithm.
Parameters: two word senses in "word#pos#sense" format.
Returns: Unless a problem occurs, the return value is the relatedness
score, which is greater-than or equal-to 0. If an error occurs,
then the error level is set to non-zero and an error
string is created (see the description of getError()).
=cut
sub getRelatedness
{
my $self = shift;
my $wps1 = shift;
my $wps2 = shift;
my $wn = $self->{wn};
my $class = ref $self || $self;
# Check the existence of the WordNet::QueryData object.
unless($wn)
{
$self->{errorString} .= "\nError (${class}::getRelatedness()) - ";
$self->{errorString} .= "A WordNet::QueryData object is required.";
$self->{error} = 2;
return undef;
}
# Using validation code from parseWps() in a super-class
my $ret = $self->parseWps($wps1, $wps2);
ref $ret or return undef;
# Initialize traces.
$self->{traceString} = "";
# Now check if the similarity value for these two synsets is in
# fact in the cache... if so return the cached value.
my $relatedness =
$self->{doCache} ? $self->fetchFromCache ($wps1, $wps2) : undef;
defined $relatedness and return $relatedness;
# Now get down to really finding the relatedness of these two.
# see if any traces reqd. if so, put in the synset arrays.
if($self->{trace})
{
# ah so we do need SOME traces! put in the synset names.
$self->{traceString} = "Synset 1: $wps1\n";
$self->{traceString} .= "Synset 2: $wps2\n";
}
# NOTE: Thanks to Wybo Wiersma for contributing optimizations
# in the following code.
# Get the gloss strings from the get_wn_info module
my ($firstStringArray, $secondStringArray, $weightsArray, $functionsStringArray) = $self->getSuperGlosses($wps1, $wps2);
my $score = 0;
for(my $i = 0; $i < scalar(@{$weightsArray}); $i++)
{
my $functionsScore = 0;
my $funcStringPrinted = 0;
# so those are the two strings for this relation pair. get the
# string overlaps
my ($overlaps, $wc1, $wc2);
if(defined($firstStringArray->[$i]) && defined($secondStringArray->[$i])
&& $firstStringArray->[$i] ne "" && $secondStringArray->[$i] ne "")
{
($overlaps, $wc1, $wc2) = $self->{finder}->getOverlaps($firstStringArray->[$i], $secondStringArray->[$i]);
}
my $overlapsTraceString = "";
my $key;
foreach $key (keys %{$overlaps})
{
# find the length of the key, square it, multiply with its
# value and finally with the weight associated with this
# relation pair to get the score for this particular
# overlap.
my @tempArray = split(/\s+/, $key);
my $value = ($#tempArray + 1) * ($#tempArray + 1) * $overlaps->{$key};
$functionsScore += $value;
# put this overlap into the trace string, if necessary
if($self->{trace} == 1)
{
$overlapsTraceString .= "$overlaps->{$key} x \"$key\" ";
}
}
# normalize the function score computed above if required
if ($self->{normalize} && defined($wc1) && defined($wc2) && ($wc1 * $wc2))
{
$functionsScore /= $wc1 * $wc2;
}
# weight functionsScore with weight of this function
$functionsScore *= $weightsArray->[$i];
# add to main score for this sense
$score += $functionsScore;
# if we have an overlap, send functionsString, functionsScore
# and overlapsTraceString to trace string, if trace string requested
if($self->{trace} == 1 && $overlapsTraceString ne "")
{
$self->{traceString} .= "".($functionsStringArray->[$i]).": $functionsScore\n";
$funcStringPrinted = 1;
$self->{traceString} .= "Overlaps: $overlapsTraceString\n";
}
# check if the two strings need to be reported in the trace.
if ($self->{trace} == 2)
{
if(!$funcStringPrinted)
{
$self->{traceString} .= "".($functionsStringArray->[$i])."\n";
$funcStringPrinted = 1;
}
$self->{traceString} .= "String 1: \"".($firstStringArray->[$i])."\"\n";
$self->{traceString} .= "String 2: \"".($secondStringArray->[$i])."\"\n";
}
}
# that does all the scoring. Put in cache if doing caching. Then
# return the score.
$self->{doCache} and $self->storeToCache($wps1, $wps2, $score);
return $score;
}
1;
__END__
=back
=head2 Usage
The semantic relatedness modules in this distribution are built as classes
that define the following methods:
new()
getRelatedness()
getError()
getTraceString()
See the WordNet::Similarity(3) documentation for details of these methods.
=head3 Typical Usage Examples
To create an object of the lesk measure, we would have the following
lines of code in the Perl program.
use WordNet::Similarity::lesk;
$measure = WordNet::Similarity::lesk->new($wn, '/home/sid/lesk.conf');
The reference of the initialized object is stored in the scalar variable
'$measure'. '$wn' contains a WordNet::QueryData object that should have been
created earlier in the program. The second parameter to the 'new' method is
the path of the configuration file for the lesk measure. If the 'new'
method is unable to create the object, '$measure' would be undefined. This,
as well as any other error/warning may be tested.
die "Unable to create object.\n" if(!defined $measure);
($err, $errString) = $measure->getError();
die $errString."\n" if($err);
To find the semantic relatedness of the first sense of the noun 'car' and
the second sense of the noun 'bus' using the measure, we would write
the following piece of code:
$relatedness = $measure->getRelatedness('car#n#1', 'bus#n#2');
To get traces for the above computation:
print $measure->getTraceString();
However, traces must be enabled using configuration files. By default
traces are turned off.
=head1 CONFIGURATION FILE
The behavior of the measures of semantic relatedness can be controlled by
using configuration files. These configuration files specify how certain
parameters are initialized within the object. A configuration file may be
specified as a parameter during the creation of an object using the new
method. The configuration files must follow a fixed format.
Every configuration file starts with the name of the module ON THE FIRST LINE
of the file. For example, a configuration file for the lesk module will have
on the first line 'WordNet::Similarity::lesk'. This is followed by the various
parameters, each on a new line and having the form 'name::value'. The
'value' of a parameter is optional (in case of boolean parameters). In case
'value' is omitted, we would have just 'name::' on that line. Comments are
supported in the configuration file. Anything following a '#' is ignored till
the end of the line.
The module parses the configuration file and recognizes the following
parameters:
=over
=item trace
The value of this parameter specifies the level of tracing that should
be employed for generating the traces. This value
is an integer equal to 0, 1, or 2. If the value is omitted, then the
default value, 0, is used. A value of 0 switches tracing off. A value
of 1 or 2 switches tracing on. A value of 1 displays as
traces only the gloss overlaps found. A value of 2 displays as traces all
the text being compared.
=item cache
The value of this parameter specifies whether or not caching of the
relatedness values should be performed. This value is an
integer equal to 0 or 1. If the value is omitted, then the default
value, 1, is used. A value of 0 switches caching 'off', and
a value of 1 switches caching 'on'.
=item maxCacheSize
The value of this parameter indicates the size of the cache, used for
storing the computed relatedness value. The specified value must be
a non-negative integer. If the value is omitted, then the default
value, 5,000, is used. Setting maxCacheSize to zero has
the same effect as setting cache to zero, but setting cache to zero is
likely to be more efficient. Caching and tracing at the same time can result
in excessive memory usage because the trace strings are also cached. If
you intend to perform a large number of relatedness queries, then you
might want to turn tracing off.
=item relation
The value of this parameter is the path to a file that contains a list of
WordNet relations. The path may be either an absolute path or a relative
path.
The lesk measure combines glosses of synsets related to the target
synsets by these relations and then searches for overlaps in these
"super-glosses."
WARNING: the format of the relation file is different for the vector and lesk
measures.
=item stop
The value of this parameter the path of a file containing a list of stop
words that should be ignored in the glosses. The path may be either an
absolute path or a relative path.
=item stem
The value of this parameter indicates whether or not stemming should be
performed. The value must be an integer equal to 0 or 1. If the
value is omitted, then the default value, 0, is used.
A value of 1 switches 'on' stemming, and a value of 0 switches stemming
'off'. When stemming is enabled, all the words of the
glosses are stemmed before their vectors are created for the vector
measure or their overlaps are compared for the lesk measure.
=item normalize
The value of this parameter indicates whether or not normalization of
scores is performed. The value must be an integer equal to 0 or 1. If
the value is omitted, then the default value, 0, is assumed. A value of
1 switches 'on' normalizing of the score, and a value of 0 switches
normalizing 'off'. When normalizing is enabled, the score obtained by
counting the gloss overlaps is normalized by the size of the glosses.
The details are described in Banerjee and Pedersen (2002).
=back
=head1 RELATION FILE FORMAT
The relation file starts with the string "RelationFile" on the first line
of the file. Following this, on each consecutive line, a relation is specified
in the form --
func(func(func... (func)...))-func(func(func... (func)...)) [weight]
Where "func" can be any one of the following functions:
hype() = Hypernym of
hypo() = Hyponym of
holo() = Holonym of
mero() = Meronym of
attr() = Attribute of
also() = Also see
sim() = Similar
enta() = Entails
caus() = Causes
part() = Particle
pert() = Pertainym of
glos = gloss (without example)
example = example (from the gloss)
glosexample = gloss + example
syns = synset of the concept
Each of these specifies a WordNet relation. And the outermost function in the
nesting can only be one of glos, example, glosexample or syns. The set of
functions to the left of the "-" are applied to the first word sense. The
functions to the right of the "-" are applied to the second word sense. An
optional weight can be specified to weigh the contribution of that relation
in the overall score.
For example,
glos(hype(hypo))-example(hype) 0.5
means that the gloss of the hypernym of the hyponym of the first synset is
overlapped with the example of the hypernym of the second synset to get the
lesk score. This score is weighted 0.5. If "glos", "example", "glosexample"
or "syns" is not provided as the outermost function of the nesting, the
measure assumes "glos" as the default.
So,
glos(hypo(also))-glos(holo(attr))
and
hypo(also)-holo(attr)
are treated the same by the measure.
=head1 SEE ALSO
perl(1), WordNet::Similarity(3), WordNet::QueryData(3)
http://www.cs.utah.edu/~sidd
http://wordnet.princeton.edu
http://www.ai.mit.edu/~jrennie/WordNet
http://groups.yahoo.com/group/wn-similarity
=head1 AUTHORS
Ted Pedersen, University of Minnesota Duluth
tpederse at d.umn.edu
Satanjeev Banerjee, Carnegie Mellon University, Pittsburgh
banerjee+ at cs.cmu.edu
Siddharth Patwardhan, University of Utah, Salt Lake City
sidd at cs.utah.edu
=head1 BUGS
None.
To report bugs, go to http://groups.yahoo.com/group/wn-similarity/ or
e-mail "S<tpederse at d.umn.edu>".
=head1 COPYRIGHT AND LICENSE
Copyright (c) 2005, Ted Pedersen, Satanjeev Banerjee and Siddharth Patwardhan
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to
The Free Software Foundation, Inc.,
59 Temple Place - Suite 330,
Boston, MA 02111-1307, USA.
Note: a copy of the GNU General Public License is available on the web
at L<http://www.gnu.org/licenses/gpl.txt> and is included in this
distribution as GPL.txt.
=cut