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.
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";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.
USAGE
The semantic relatedness modules in this distribution are built as classes that expose the following methods: new() getRelatedness() getError() getTraceString()
See the WordNet::Similarity(3) documentation for details of these methods.
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.
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:
(a) 'trace::' -- The value of this parameter specifies the level of tracing that should be employed for generating the traces. This value is an integer 0, 1 or 2. A value of 0 switches tracing off. A value of 1 displays as traces only the gloss overlaps found. A value of 2 displays as traces, all the text being compared.
(b) 'cache::' -- can take values 0 or 1 or the value can be omitted, in which case it takes the value 1, i.e. switches 'on' caching. A value of 0 switches caching 'off'. By default caching is enabled.
(c) 'relation::' -- The value is a filename (with complete path) of a file that contains a list of WordNet-relations. The vector module combines the glosses of synsets related to the target synsets by these relations, and forms the gloss-vector from this combined gloss. The format of the relation file is specified later in the documentation.
(d) 'stop::' -- The value is a string that specifies the path of a file containing a list of stop words that should be ignored for the gloss overlaps.
(e) 'stem::' -- can take values 0 or 1 or the value can be omitted, in which case it takes the value 1, i.e. switches 'on' stemming. A value of 0 switches stemming 'off'. When stemming is enabled, all the words of the glosses are stemmed before their overlaps are determined.
(f) 'normalize::' -- can take values 0 or 1 or the value can be omitted, in which case it takes the value 1, i.e. switches 'on' normalizing of the score. 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 Pedersen (2002).
(g) 'maxCacheSize::' -- takes a non-negative integer value. The value indicates the size of the cache, used for storing the computed relatedness value.
RELATION FILE FORMAT
The relation file starts with the string "LeskRelationFile" 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.
SEE ALSO
perl(1), WordNet::Similarity(3), WordNet::QueryData(3)
http://www.cs.utah.edu/~sidd
http://www.cogsci.princeton.edu/~wn
http://www.ai.mit.edu/~jrennie/WordNet
http://groups.yahoo.com/group/wn-similarity
AUTHORS
Satanjeev Banerjee, <banerjee+@cs.cmu.edu>
Ted Pedersen, <tpederse@d.umn.edu>
Siddharth Patwardhan, <sidd@cs.utah.edu>COPYRIGHT AND LICENSE
Copyright 2003 by Satanjeev Banerjee, Ted Pedersen and Siddharth Patwardhan
This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.