The impact of machine translation error types on post-editing effort indicators

Publication type
C1
Publication status
Published
Authors
Daems, J, Vandepitte, S., Hartsuiker, R., & Macken, L.
Editor
Sharon O'Brien and Michel Simard
Series
Fourth Workshop on Post-Editing Technology and Practice, Proceedings
Pagination
31-45
Publisher
Association for Machine Translation in the Americas
Conference
4th Workshop on Post-Editing Technology and Practice (WPTP4) (Miami)
Download
(.pdf)
Project
Robot
View in Biblio
(externe link)

Abstract

In this paper, we report on a post-editing study for general text types from English into Dutch conducted with master's students of translation. We used a fine-grained machine translation (MT) quality assessment method with error weights that correspond to severity levels and are related to cognitive load. Linear mixed effects models are applied to analyze the impact of MT quality on potential post-editing effort indicators. The impact of MT quality is evaluated on three different levels, each with an increasing granularity. We find that MT quality is a significant predictor of all different types of post-editing effort indicators and that different types of MT errors predict different post-editing effort indicators.