NAME
   README - [documentation] Introduction to WordNet::Similarity

SYNOPSIS
   There are a number of documentation files covering different aspects of
   WordNet::Similarity available:

   *   intro.pod Introduction to WordNet::Similarity (this document)

   *   install.pod How to install

   *   utils.pod How to use the utility programs in /utls

   *   modules.pod How the measure modules are designed

   *   developers.pod How to write your own measure of semantic relatedness

   *   config.pod How to set configuration options for the measures

   You can use pod2html, pod2latex, pod2man, or pod2text to get this
   documentation in a different format. See the man pages for pod2html,
   etc. These translators should come with Perl, but you can also download
   them from <http://search.cpan.org>.

DESCRIPTION
   This package consists of Perl modules along with supporting Perl
   programs that implement the semantic relatedness measures described by
   Leacock & Chodorow (1998), Jiang & Conrath (1997), Resnik (1995), Lin
   (1998), Hirst & St-Onge (1998), Wu & Palmer (1994), the extended gloss
   overlap measure by Banerjee and Pedersen (2002), and two measures based
   on context vectors by Patwardhan (2003). The details of the vector
   measure are described in the Master's thesis work of Patwardhan (2003)
   at the University of Minnesota Duluth, and the vector_pairs measure is
   derived from that.

   The Perl modules are designed as objects with methods that take as input
   two word senses. The semantic relatedness of these word senses is
   returned by these methods. A quantitative measure of the degree to which
   two word senses are related has wide ranging applications in numerous
   areas, such as word sense disambiguation, information retrieval, etc.
   For example, in order to determine which sense of a given word is being
   used in a particular context, the sense having the highest relatedness
   with its context word senses is most likely to be the sense being used.
   Similarly, in information retrieval, retrieving documents containing
   highly related concepts are more likely to have higher precision and
   recall values.

   A command line interface to these modules is also present in the
   package. The simple, user-friendly interface returns the relatedness
   measure of two given words. A number of switches and options have been
   provided to modify the output and enhance it with trace information and
   other useful output. Details of the usage are provided in other sections
   of this README. Supporting utilities for generating information content
   files from various corpora are also available in the package. The
   information content files are required by three of the measures for
   computing the relatedness of concepts.

   The following sections describe the organization of this software
   package and how to use it. A few typical examples are given to help
   clearly understand the usage of the modules and the supporting
   utilities.

SEMANTIC RELATEDNESS
   We observe that humans find it extremely easy to say if two words are
   related and if one word is more related to a given word than another.
   For example, if we come across two words -- 'car' and 'bicycle', we know
   they are related as both are means of transport. Also, we easily observe
   that 'bicycle' is more related to 'car' than 'fork' is. But is there
   some way to assign a quantitative value to this relatedness? Some ideas
   have been put forth by researchers to quantify the concept of
   relatedness of words, with encouraging results.

   A number of different measures of relatedness have been implemented in
   this software package. These include a simple edge counting approach and
   a random method for measuring relatedness. The measures rely heavily on
   the vast store of knowledge available in the online electronic
   dictionary -- WordNet. So, we use a Perl interface for WordNet called
   WordNet::QueryData to make it easier for us to access WordNet. The
   modules in this package REQUIRE that the WordNet::QueryData module be
   installed on the system before these modules are installed.

CONTENTS
   The package contains the semantic relatedness modules, some support Perl
   utilities and some sample configuration files, data files and programs.

 Modules
   All the modules that will be installed in the Perl system directory are
   present in the '/lib' directory tree of the package. These include the
   semantic relatedness modules -- jcn.pm, res.pm, lin.pm, lch.pm, hso.pm,
   lesk.pm, vector.pm, vector_pairs.pm, wup.pm, path.pm and random.pm --
   present in the /lib/WordNet/Similarity subdirectory and the supporting
   modules get_wn_info.pm and string_compare.pm. There also exists a
   WordNet/Similarity.pm module that currently serves as a base class for
   all the relatedness modules and contains Perl documentation. All these
   modules, once installed in the Perl system directory, can be directly
   used by Perl programs.

 Supporting Utilities
   The '/utils' subdirectory of the package contains supporting Perl
   programs. 'similarity.pl' is a command-line interface to the relatedness
   modules. A number of Perl programs, that generate information content
   files from various corpora, are provided. As part of the standard
   install, these are also installed into the system directories, and can
   be accessed from any working directory if the common system directories
   (/usr/bin, /usr/local/bin, etc) are in your path.

 Samples
   If you downloaded this package as a tar-gzipped file from the web, you
   will find a '/samples' subdirectory in the package. There is a separate
   README in that directory. The directory contains sample configuration
   files for the modules, sample programs showing usage of the modules and
   sample data files (e.g., relation files).

REFERENCES
   1   Wu Z. and Palmer M. 1994. Verb Semantics and Lexical Selection. In
       Proceedings of the 32nd Annual Meeting of the Association for
       Computational Linguistics. Las Cruces, New Mexico.

   2   Resnik P. 1995. Using information content to evaluate semantic
       similarity. In Proceedings of the 14th International Joint
       Conference on Artificial Intelligence, pages 448-453. Montreal.

   3   Jiang J. and Conrath D. 1997. Semantic similarity based on corpus
       statistics and lexical taxonomy. In Proceedings of International
       Conference on Research in Computational Linguistics. Taiwan.

   4   Fellbaum C., editor. WordNet: An electronic lexical database. MIT
       Press, 1998.

   5   Leacock C. and Chodorow M. 1998. Combining local context and WordNet
       similarity for word sense identification. In Fellbaum 1998, pp.
       265-283.

   6   Lin D. 1998. An information-theoretic definition of similarity. In
       Proceedings of the 15th International Conference on Machine
       Learning. Madison, WI.

   7   Hirst G. and St-Onge D. 1998. Lexical Chains as representations of
       context for the detection and correction of malapropisms. In
       Fellbaum 1998, pp. 305-332.

   8   Sch�tze H. 1998. Automatic Word Sense Discrimination. Computational
       Linguistics, 24(1):97-123.

   9   Resnik P. 1999. Semantic Similarity in a Taxonomy: An Information-
       Based Measure and its Applications to Problems of Ambiguity in
       Natural Language. Journal of Artificial Intelligence Research, 11,
       95-130.

   10  Budanitsky A. and Hirst G. 2001. Semantic distance in WordNet: An
       experimental, application-oriented evaluation of five measures. In
       Workshop on WordNet and Other Lexical Resources, Second meeting of
       the North American Chapter of the Association for Computational
       Linguistics. Pittsburgh, PA.

   11  Banerjee S. and Pedersen T. 2002. An Adapted Lesk Algorithm for Word
       Sense Disambiguation Using WordNet. In Proceeding of the Fourth
       International Conference on Computational Linguistics and
       Intelligent Text Processing (CICLING-02). Mexico City.

   12  Patwardhan S., Banerjee S. and Pedersen T. 2002. Using Semantic
       Relatedness for Word Sense Disambiguation. In Proceedings of the
       Fourth International Conference on Intelligent Text Processing and
       Computational Linguistics. Mexico City.

   13  Banerjee S. and Pedersen T. 2003. Extended Gloss Overlaps as a
       Measure of Semantic Relatedness. In the Proceedings of the
       Eighteenth International Joint Conference on Artificial
       Intelligence. Acapulco, Mexico.

   14  Patwardhan S. and Pedersen T. 2006. Using WordNet-based Context
       Vectors to Estimate the Semantic Relatedness of Concepts. In the
       Proceedings of the EACL 2006 Workshop Making Sense of Sense -
       Bringing Computational Linguistics and Psycholinguistics Together.
       Trento, Italy.

   15  Banerjee S. Adapting the Lesk algorithm for word sense
       disambiguation to WordNet. Master Thesis, University of Minnesota,
       Duluth, 2002.

   16  Patwardhan S. Incorporating dictionary and corpus information into a
       vector measure of semantic relatedness. Master Thesis, University of
       Minnesota, Duluth, 2003.

SEE ALSO
   intro.pod

   Mailing list: <http://groups.yahoo.com/group/wn-similarity>

   Project Home page: <http://wn-similarity.sourceforge.net>

ACKNOWLEDGEMENTS
   We would like to thank the following for their support and contribution
   towards the development of this package. We thank Jason Rennie for his
   QueryData package, the WordNet guys at Princeton for WordNet, Resnik,
   Hirst, St-Onge, Jiang, Conrath, Lin, Wu, Palmer, Leacock, and Chodorow
   for their algorithms and work on the relatedness measures. We also thank
   Bano (Satanjeev Banerjee) for his work on the extended gloss overlap
   module. We are grateful to Wybo Wiersma for contributing his
   optimizations to the GlossFinder code. We also appreciate the many
   helpful suggestions and bug patches from Ben Haskell.

AUTHORS
    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

    Jason Michelizzi

COPYRIGHT
   Copyright (c) 2005-2008, Ted Pedersen, Siddharth Patwardhan, Satanjeev
   Banerjee, and Jason Michelizzi

   Permission is granted to copy, distribute and/or modify this document
   under the terms of the GNU Free Documentation License, Version 1.2 or
   any later version published by the Free Software Foundation; with no
   Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts.

   Note: a copy of the GNU Free Documentation License is available on the
   web at <http://www.gnu.org/copyleft/fdl.html> and is included in this
   distribution as FDL.txt.

You are viewing proxied material from ftp.icm.edu.pl. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.