Direction matters: Comparing post-editing and human translation effort and quality.

Area of Science:

• Computational Linguistics
• Human-Computer Interaction
• Translation Studies

Background:

• Previous research has compared post-editing (PE) and human translation (HT) but rarely considered the impact of translation direction.
• The interplay between translation mode (PE vs. HT) and directionality (L1-L2 vs. L2-L1), especially for distant languages, is understudied.

Purpose of the Study:

• To investigate how translation direction influences neural machine translation post-editing (PE) and human translation (HT) processes.
• To analyze the effects of L1-L2 (Chinese-English) versus L2-L1 (English-Chinese) directionality on translation performance and perceived difficulty.

Main Methods:

• Thirty-one native Chinese translators performed both PE and HT tasks in both Chinese-English and English-Chinese directions.
• Translator processes were meticulously documented using keylogging and screen recording.

Main Results:

• Post-editing (PE) generally surpassed human translation (HT) in speed, efficiency, and quality, with PE advantages being more significant in the L2-L1 (English-Chinese) direction.
• Perceived task difficulty was predominantly influenced by translation direction, with L1-L2 (Chinese-English) consistently rated as more challenging irrespective of the translation mode.
• The study challenges the notion that PE universally reduces effort compared to HT, highlighting a complex interaction between mode and directionality.

Conclusions:

• Translation directionality significantly modulates the performance benefits of post-editing (PE) over human translation (HT).
• Task difficulty in translation is primarily a function of language direction, not solely the translation mode.
• Findings inform translator training, workflow optimization, and compensation models, emphasizing the need to consider both translation mode and direction.