A reordering tool for machine translation.
C++ Shell Perl Other
Switch branches/tags
Nothing to show
Permalink
Failed to load latest commit information.
example
m4
script
src
.gitignore
AUTHORS
COPYING
ChangeLog
INSTALL
Makefile.am
NEWS
README
README.fastertrain
README.paralleltrain
configure.ac
eclipse-1.0.txt

README

###########################################################################
# lader - Latent Derivation Reorderer                                     #
#   by Graham Neubig                                                      #
#   http://www.phontron.com/lader                                         #
###########################################################################

This is a parser/reorderer for machine translation.
It works by reordering source language sentences to target language order, which
allows for standard phrase-based statistical machine translation techniques
to work with only local reordering, greatly improving results.
For more information on the technique, you can reference the following paper:

Graham Neubig, Taro Watanabe, Shinsuke Mori
Inducing a Discriminative Parser to Optimize Machine Translation Reordering
EMNLP 2012
http://aclweb.org/anthology-new/D/D12/D12-1077.pdf

*** Installing ***

lader requires that you have boost installed to work properly.
Once your environment is set up, you can use the following standard commands:

$ autoreconf -i
$ ./configure [--prefix=/my/favorite/dir]
$ make
$ make install

*** Training/Running ***

You can find more information about training or running models on the lader
web site, or by following the scripts in the "example" directory.

*** Details/Acknowledgement ***
This software was partly developed while the author was an intern at NICT, Japan, and the author is grateful for being allowed the opportunity to make the code open-source.

This software is Copyright (C) by Graham Neubig <neubig@gmail.com> and National Institute of Information and Communications Technology, and is licensed under the terms of the Eclipse Public License, Version 1.0 or (at your option) any subsequent version.