WAT

The Workshop on Asian Translation

Automatic Evaluation Procedures

Tools

For Evaluation

• Moses version 2.1.1
(Using mosesdecoder/scripts/generic/multi-bleu.perl)
• RIBES version 1.02.4
• Adequacy-Fluency Metrics (AMFM) [pdf]

For Japanese segmentation

• Juman version 7.0
• KyTea 0.4.6
(Using Full SVM model (jp-0.4.2-utf8-1.mod))
• MeCab 0.996 (in Japanese)

For Chinese segmentation

• KyTea 0.4.6
(Using Full SVM Model in MSR model (Simplified) (msr-0.4.0-1.mod))
• Stanford Word Segmenter version 2014-06-16
  (Using Chinese Penn Treebank (CTB) and Peking University (PKU) model)

For English and Indonesian segmentation

• Moses version 2.1.1
(Using mosesdecoder/scripts/tokenizer/tokenizer.perl)

For Korean segmentation

• mecab-ko
  (https://bitbucket.org/eunjeon/mecab-ko/downloads/mecab-0.996-ko-0.9.1.tar.gz)
  (https://bitbucket.org/eunjeon/mecab-ko-dic/downloads/mecab-ko-dic-1.6.1-20140814.tar.gz)
  Install procedure
  

For Hindi segmentation

• Indic NLP Library version 0.3: For tokenization, normalization and script conversion

For all segmentation

• scripts

Java

• Version 7 Update 60 (JDK, JRE)
  (For Stanford Word Segmenter)

Perl

• perl 5 version 24 subversion 0 (v5.24.0)
  (The following link (https://github.com/hassyGo/WAT-scripts) helps you to use other perl enviroments.)

JST (Japan Science and Technology Agency)
NICT (National Institute of Information and Communications Technology)
Kyoto University
Last Modified: 2018-08-30