Bitext word alignment

<h2 id="training">Training</h2>
<h3>IBM Models</h3>
<p class="note">Main article: <a href="/facts/IBM_alignment_models/hYfsl0K7">IBM alignment models</a></p>
<p>The IBM models<a class="footnote-ref" id="fnref:4" href="#fn:4"><sup>4</sup></a> are used in <a href="/facts/Statistical_machine_translation/rzJ8Gfyj">Statistical machine translation</a> to train a translation model and an alignment model. They are an instance of the <a href="/facts/Expectation%E2%80%93maximization_algorithm/32P2bdBc">Expectation–maximization algorithm</a>: in the expectation-step the translation probabilities within each sentence are computed, in the maximization step they are accumulated to global translation probabilities.
Features:
</p>
<ul><li>IBM Model 1: lexical alignment probabilities</li>
<li>IBM Model 2: absolute positions</li>
<li>IBM Model 3: fertilities (supports insertions)</li>
<li>IBM Model 4: relative positions</li>
<li>IBM Model 5: fixes deficiencies (ensures that no two words can be aligned to the same position)</li></ul>
<h3>HMM</h3>
<p>Vogel et al.<a class="footnote-ref" id="fnref:5" href="#fn:5"><sup>5</sup></a> developed an approach featuring lexical translation probabilities and relative alignment by mapping the problem to a <a href="/facts/Hidden_Markov_model/ur1zTAhP">Hidden Markov model</a>. The states and observations represent the source and target words respectively. The transition probabilities model the alignment probabilities. In training the translation and alignment probabilities can be obtained from 
  
    
      
        
          γ
          
            t
          
        
        (
        i
        )
      
    
    {\displaystyle \gamma _{t}(i)}
  
 and 
  
    
      
        
          ξ
          
            t
          
        
        (
        i
        ,
        j
        )
      
    
    {\displaystyle \xi _{t}(i,j)}
  
 in the <a href="/facts/Forward-backward_algorithm/3tVWYLSH">Forward-backward algorithm</a>.
</p>
<h2 id="software">Software</h2>
<ul><li><a href="https://github.com/moses-smt/giza-pp">GIZA++</a> (free software under GPL)
<ul><li>The most widely used alignment toolkit, implementing the famous IBM models with a variety of improvements</li></ul></li>
<li><a href="https://github.com/mhajiloo/berkeleyaligner">The Berkeley Word Aligner</a> (free software under GPL)
<ul><li>Another widely used aligner implementing alignment by agreement, and discriminative models for alignment</li></ul></li>
<li><a href="https://jasonriesa.github.io/nile/">Nile</a> (free software under GPL)
<ul><li>A supervised word aligner that is able to use syntactic information on the source and target side</li></ul></li>
<li><a href="http://www.phontron.com/pialign/">pialign</a> (free software under the Common Public License)
<ul><li>An aligner that aligns both words and phrases using Bayesian learning and inversion transduction grammars</li></ul></li>
<li><a href="http://linguateca.di.uminho.pt/natools/">Natura Alignment Tools</a> (NATools, free software under GPL)</li>
<li><a href="http://research.variancia.com/unl-aligner">UNL aligner</a> (free software under Creative Commons Attribution 3.0 Unported License)</li>
<li><a href="http://nlp.cs.nyu.edu/GMA/">Geometric Mapping and Alignment (GMA)</a> (free software under GPL)</li>
<li><a href="https://elrc-share.eu/repository/browse/hunalign/ade4ee46604c11e9a7e100155d0267060a4253f9b3a54dfe83f98bb9affd85a4">HunAlign</a> (free software under LGPL-2.1)</li>
<li><a href="https://anymalign.limsi.fr/">Anymalign</a> (free software under GPL)</li></ul>

<h2 id="references">References</h2>

<ol>
<li id="fn:1"><p>P. F. Brown et al. 1993. The Mathematics of Statistical Machine Translation: Parameter Estimation Archived April 24, 2009, at the Wayback Machine. Computational Linguistics, 19(2):263–311. <a href="http://acl.ldc.upenn.edu/J/J93/J93-2003.pdf" target="_blank">http://acl.ldc.upenn.edu/J/J93/J93-2003.pdf</a> <a href="#fnref:1" class="footnote-back-ref">↩</a></p></li>
<li id="fn:2"><p>Och, F.J. and Tillmann, C. and Ney, H. and others 1999, Improved alignment models for statistical machine translation, Proc. of the Joint SIGDAT Conf. on Empirical Methods in Natural Language Processing and Very Large Corpora <a href="http://www.aclweb.org/anthology/W99-0604.pdf" target="_blank">http://www.aclweb.org/anthology/W99-0604.pdf</a> <a href="#fnref:2" class="footnote-back-ref">↩</a></p></li>
<li id="fn:3"><p>ACL 2005: Building and Using Parallel Texts for Languages with Scarce Resources Archived May 9, 2009, at the Wayback Machine <a href="http://www.cse.unt.edu/~rada/wpt05/" target="_blank">http://www.cse.unt.edu/~rada/wpt05/</a> <a href="#fnref:3" class="footnote-back-ref">↩</a></p></li>
<li id="fn:4"><p>Philipp Koehn (2009). Statistical Machine Translation. Cambridge University Press. p. 86ff. ISBN 978-0521874151. Retrieved 21 October 2015. <a href="978-0521874151" target="_blank">978-0521874151</a> <a href="#fnref:4" class="footnote-back-ref">↩</a></p></li>
<li id="fn:5"><p>S. Vogel, H. Ney and C. Tillmann. 1996. HMM-based Word Alignment in Statistical Translation Archived 2018-03-02 at the Wayback Machine. In COLING ’96: The 16th International Conference on Computational Linguistics, pp. 836-841, Copenhagen, Denmark. <a href="https://aclanthology.info/pdf/C/C96/C96-2141.pdf" target="_blank">https://aclanthology.info/pdf/C/C96/C96-2141.pdf</a> <a href="#fnref:5" class="footnote-back-ref">↩</a></p></li>
</ol>

Bitext word alignment open-in-new

Bitext word alignment