Features of Accuracy Mismatch between Back-Translated Sentences and Target-Translated Sentences

Abstract

In communication using a machine translation, inaccurate translation prevents effective communication between individuals and leads to misunderstandings. Back-translation is used to check the accuracy of a sentence translated to a native language. For back-translation to be used as a method for checking the accuracy of a translated sentence, it has to satisfy the following conditions: There must be a positive correlation between the accuracy of sentences translated to an intermediate language and that of back-translated sentences, and there can be no significant difference between them. From the results of our verification, we found that back-translation satisfies these conditions. However, we found that accuracy mismatch case 1, which means that a back-translated sentence is accurate but the translated sentence is inaccurate, occurred although the number of such accuracy mismatch cases was small. Accuracy mismatch case 1 can lead to serious problems in communication. In this paper, we discuss the causes of accuracy mismatch case 1. We find that there are six features of accuracy mismatch between a back-translated sentence and its target-translated sentence.