Proj CDeepFuzz Paper Reading: Testing machine translation via referential transparency
Abstract
本文:RTIs(Referentially Transparent Inputs), Purity
Task: Testing Machine Translation Model
Method: referentially transparent input(在不同的上下文中会有相同翻译的文本）
实验：
对象：Google Translate, Bing
数据集： 200 unlabeled sentences
效果：
- detected 123 and 142 erroneous translations，包括under translation, over translation, word/phrase mistranslation, incorrect modification, unclear logic
