Review Article

Extracting Parallel Sentences from Nonparallel Corpora Using Parallel Hierarchical Attention Network

Table 1

Training and test set statistics.

TypeLanguageNumber

Training dataEnglish-French229,000
English-Chinese287,000
English-German237,000

Test dataEnglish-FrenchEnglish38,069
French21,497
English-ChineseEnglish88,860
Chinese94,637
English-GermanEnglish40,354
German32,594