apertium-tagger-training-tools

This software is GPL. It is provided as is, WITHOUT any guarantee, in
the hope that it will be useful.

(c) 2004-2006 Felipe Snchez Martnez
(c) 2006 Universitat d'Alacant
---------------------------------------------------------------------

This package provides a set of tools to train and test the
hidden-Markov-model-based part-of-speech tagger module of the apertium
machine translation engine (http://www.apertium.org).

The main program provided with this package apertium-tagger-tl-trainer
uses information from the target language to train the part-of-speech
tagger used in apertium in an unsupervised way, i. e. without human
intervention. This method has been already described (see below) and
is known to give better results than the standard Baum-Welch algorithm
used to train hidden-Markov-model-based part-of-speech taggers when no
tagged text is available.

Comments on this software are VERY welcome, do not hesitate to write
me for comments, feedback or just to tell me that you found this
software useful.

If you use this software for research, or for any other particular
purpose, I would be very grateful if you could cite one or both of the
following research papers describing this software:

* Felipe Snchez-Martnez, Juan Antonio Prez-Ortiz, Mikel
  L. Forcada. Exploring the use of target-language information to
  train the part-of-speech tagger of machine translation systems. In
  Lecture Notes in Computer Science 3230 (Advances in Natural Language
  Processing, Proceedings of EsTAL - Espaa for Natural Language
  Processing), p. 137-148, October 20-22, 2004, Alicante, Spain.

  @incollection{sanchezmartinez04,
  author    = {S{\'{a}}nchez-Mart{\'{\i}}nez, Felipe 
               and P{\'{e}}rez-Ortiz, Juan Antonio 
               and Forcada, Mikel L.},
  title     = {Exploring the use of target-language information to train the
               part-of-speech tagger of machine translation systems},
  booktitle = {Advances in Natural Language Processing, Proceedings of 
               4th International Conference {EsTAL}},
  publisher = {Springer-Verlag},
  year      = {2004},
  month     = {October},
  volume  = {3230},
  series  = {Lecture Notes in Computer Science},
  pages   = {137--148},
  address = {},
  note    = {},
  location     = {Alicante, Spain},
  copyright    = {Springer-Verlag},
  isbn         = {3-540-23498-5},
  issn         = {0302-9743},
  }


* Felipe Snchez-Martnez, Juan Antonio Prez-Ortiz, Mikel L. Forcada.
  Speeding up target-language driven part-of-speech tagger training for
  machine translation. In Lecture Notes in Computer Science 4293 (Advances in
  Artificial Intelligence, Proceedings of MICAI 2006, 5th Mexican
  International Conference on Artificial Intelligence) , p. 844-854,
  November 13-17, 2006, Apizaco, Mexico.

  @incollection{sanchezmartinez06,
  author = {S{\'{a}}nchez-Mart{\'{\i}}nez, Felipe
            and P{\'{e}}rez-Ortiz, Juan Antonio
            and Forcada, Mikel L.},
  title =  {Speeding up target-language driven part-of-speech tagger
            training for machine translation},
  booktitle = {{MICAI} 2006: {A}dvances in {A}rtificial {I}ntelligence,
               Proceedings of the 5th {M}exican {I}nternational {C}onference
               on {A}rtificial {I}ntelligence},
  year = {2006},
  publisher = {Springer-Verlag},
  volume =       {4293},
  pages =        {844--854},
  series =       {Lecture Notes in Computer Science},
  month =        {November},
  location     = {Apizaco, Mexico},
  copyright    = {Springer-Verlag},
  isbn         = {3-540-37334-9},
  issn         = {0302-9743},
  }

Contact information:
---------------------------------------------------------------------

Felipe Snchez Martnez

Departament de Llenguatges y Sistemes Informtics
Universitat d'Alacant
E-03071 Alacant (Spain)

fsanchez@dlsi.ua.es
http://www.dlsi.ua.es/~fsanchez

Installation
---------------------------------------------------------------------
To install apertium-tagger-training-tools you need to have installed
the following packages:

libxml2, at least version 2.6.17 
lttoolbox, at least version 3.0.0
apertium, at least version 3.0.0

Once you have all the required software installed, install
apertium-tagger-training-tools in the common way:

$ ./configure && make

and as a root

# make install

Good luck!

Acknowledgements
---------------------------------------------------------------------
Software funded by the Spanish Ministry of Science and Technology
through project TIC2003-08681-C02-01, and by the Spanish Ministry of
Education and Science and the European Social Found through research
grant BES-2004-4711.

References
---------------------------------------------------------------------
Please visit http://www.dlsi.ua.es/~fsanchez/publications.html

Example of use
---------------------------------------------------------------------
Have a look at the 'example' directory
