Domain-Specific Machine Translation for Software Localization

App Quality Data

There are 50 .txt files in total. Each .txt file contains text of an app, which is randomly selected for quality checking. .

Tool Usefulness

There are 4 folders (each contains 8 files for 8 randomly selected apps):

  • cn: Chinese ground truth
  • en: English source sentence
  • T1: Translation from translator 1 (translate on him/her own)
  • T2: Translation from translator 2 (provided with C-2RNN translation)

Popular App Results

Each folder contains 5 files:

  • gt: Ground truth Chinese sentences
  • en: English source sentences
  • c2rnn: Translation of C-2RNN
  • google: Translation of Google Translate
  • phrasal: Translation of phrase-based MT system

Uncommon App Results

Each folder contains 5 files:

  • gt: Ground truth Chinese sentences
  • en: English source sentences
  • c2rnn: Translation of C-2RNN
  • google: Translation of Google Translate
  • phrasal: Translation of phrase-based MT system