++ed by:

4 non-PAUSE users.

Author image Ted Pedersen
and 1 contributors


WordNet::Similarity::vector_pairs - module for computing semantic relatedness of word senses using second order co-occurrence vectors of glosses of the word senses.


  use WordNet::Similarity::vector_pairs;

  use WordNet::QueryData;

  my $wn = WordNet::QueryData->new();

  my $vector_pairs = WordNet::Similarity::vector_pairs->new($wn);

  my $value = $vector_pairs->getRelatedness("car#n#1", "bus#n#2");

  ($error, $errorString) = $vector_pairs->getError();

  die "$errorString\n" if($error);

  print "car (sense 1) <-> bus (sense 2) = $value\n";


Schütze (1998) creates what he calls context vectors (second order co-occurrence vectors) of pieces of text for the purpose of Word Sense Discrimination. This idea is adopted by Patwardhan and Pedersen to represent the word senses by second-order co-occurrence vectors of their dictionary (WordNet) definitions. The relatedness of two senses is then computed as the cosine of their representative gloss vectors.

A concept is represented by its own gloss, as well as the glosses of the neighboring senses as specified in the vector-relation.dat file. Each gloss is converted into a second order vector by replacing the words in the gloss with co-occurrence vectors for those words. The overall measure of relatedness between two concepts is determined by taking the pairwise cosines between these expanded glosses. If vector-relation.dat consists of:


then three pairwise cosine measurements are made to determine the relatedness of concepts A and B. The examples found in the glosses of A and B are expanded and measured, then the glosses themselves are expanded and measured, and then the hyponyms of A and B are expanded and measured. Then, the values of these three pairwise measures are summed to create the overall relatedness score.


Overrides the initialize method in the parent class (GlossFinder.pm). This method essentially initializes the measure for use.

Parameters: $file -- configuration file.

Returns: none.


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.


Computes the relatedness of two word senses using the Vector 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()).


The semantic relatedness modules in this distribution are built as classes that define the following methods:


See the WordNet::Similarity(3) documentation for details of these methods.

Typical Usage Examples

To create an object of the vector_pairs measure, we would have the following lines of code in the Perl program.

  use WordNet::Similarity::vector_pairs;
  $measure = WordNet::Similarity::vector_pairs->new($wn, '/home/sid/vector_pairs.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 vector_pairs 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.


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 vector_pairs module will have on the first line 'WordNet::Similarity::vector_pairs'. 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:


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.


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'.


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.


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 vector_pairs module combines the glosses of synsets related to the target synsets by these relations and forms the gloss-vector from this combined gloss.

WARNING: the format of the relation file is different for the vector_pairs and lesk measures.


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.


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.


The value of this parameter is the path to a file containing word vectors, i.e. co-occurrence vectors for all the words in the WordNet glosses. The value of this parameter may not be omitted, and the vector_pairs measure will not run without a vectors file being specified in a configuration file.


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 = the 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 functions specify which glosses to use for forming the gloss vector of the synset. An optional weight can be specified to weigh the contribution of that relation in the overall score.

For example,

 glos(hype(hypo))-glosexample(hype) 0.5

means that the gloss of the hypernym of the hyponym of the first synset is used to form the gloss vector of the first synset, and the gloss+example of the hypernym of the second synset is used to form the gloss vector of the second synset. The values in these vector are weighted by 0.5. If one of "glos", "example", "glosexample" or "syns" is not specified as the outermost function in the nesting, then "glosexample" is assumed by default. This implies that




are equivalent as far as the measure is concerned.


perl(1), WordNet::Similarity(3), WordNet::QueryData(3)






 Ted Pedersen, University of Minnesota, Duluth
 tpederse at d.umn.edu

 Siddharth Patwardhan, University of Utah, Salt Lake City
 sidd at cs.utah.edu

 Satanjeev Banerjee, Carnegie Mellon University, Pittsburgh
 banerjee+ at cs.cmu.edu


To report bugs, go to http://groups.yahoo.com/group/wn-similarity/ or send an e-mail to "tpederse at d.umn.edu".


Copyright (c) 2005, Ted Pedersen, Siddharth Patwardhan and Satanjeev Banerjee

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 http://www.gnu.org/licenses/gpl.txt and is included in this distribution as GPL.txt.