!!!Sentence alignment

We look at alternatives to our [tca2|tca2.html]  aligners.

!!!The Europarl aligner

The sentence aligner used to align [the Europarl parallel corpus|http://www.statmt.org/europarl/] is a perl script based upon [Church and Gale algorithm|http://people.csail.mit.edu/koehn/publications/de-news/church_and_gale.ps]. Here is [the README file|europarl_aligner_README.txt].


The script may be downloaded from the europarl site. At Giellatekno, it is placed under the {{$GTHOME/tools/alignment-tools/europarl/}}.

In order to run it: Add se and no abbreviation files to the ''nonbreaking_prefixes'' catalogue (this has been done). Then add files to the directories se and no. The filenames in se and no must be ''identical''. The command then is

./sentence-align-corpus.perl se no

This might not work. For some hints, see:

* [Struggeling with documentation|http://www.mail-archive.com/moses-support@mit.edu/msg00301.html]



__TODO:__

* Run a test on the Europarl aligner, and compare to tca2.


!!!Other aligners?

Feel free to add.