concordia-server/mgiza-aligner/europarl/README
2017-01-21 17:01:15 +01:00

59 lines
1.8 KiB
Plaintext

Europarl Release v3 -- Sept 27, 2007
===================================
This is a parallel corpus that was extracted from the
European Parliament web site by Philipp Koehn (University
of Edinburgh). It is faily big, 40 million words per
language, and its main intended use is to aid
statistical machine translation research.
More information can be found at
http://www.statmt.org/europarl/
The main difference in this release vs. the first release
in 2002 and second release in 2003 is that it is larger
and it comes with improved processing tools that allow
the creation of parallel corpora between any two of the
11 languages.
Some data is now tagged with the original language the text
was spoken in.
Sentence aligner
----------------
You can create any parallel corpus with the command
./sentence-align-corpus.perl L1 L2
where L1 and L2 can be any of the 11 languages
da de el en es fi fr it nl pt sv
The output is stored in the aligned/ directory.
NOTE: To use this corpus with tools like Giza++, you want to
- lowercase the text (recommended)
- strip empty lines and their correspondences (recommended)
- tokenize words and punctuation (recommended)
- remove lines with XML-Tags (starting with "<") (required)
The sentence aligner uses the split-sentences.perl script,
which does and sentence splitting. You may want to
use your own preprocessor. This requires changing an
obvious line in the sentence aligner code. A tokenizer.perl
script is included as well.
Source
------
http://www3.europarl.eu.int/omk/omnsapir.so/calendar?APP=CRE&LANGUE=EN
Copyright in the Europarl service
(c) European Communities
Except where otherwise indicated, reproduction is authorised,
provided that the source is acknowledged.
Change Log
----------
Preprocessing is improved.
This release covers 9/1996 - 10/2006.
Includes sentence aligner and tokenizer.