Big Wrong Data: Insights from developing and using an automatic translation revision tool with error memories

Abstract

Since August 2017, more than 40 institutions (and over 800 translators) piloted the translationQ revision platform. This paper will not focus on the development of but shows some lessons learned from translationQ’s error/revision memories: how to recycle the errors (‘Big Wrong Data’) into useful insights for translation, translation evaluation and translation teaching. The translationQ project (a joint initiative of KU Leuven and Televic Education) was developed to automate and speed up evaluation processes in both translator education and the translation profession. The tool works for bilingual or monolingual text productions. The program’s core is an error or revision ‘memory’: it allows the system to recognize errors in new translations and to suggest corrections and feedback automatically; the program leaves room for human intervention. Still, the bulk workload of retyping the same corrections and feedback time and again is now done automatically by the tool, leading to more rapid and more consistent revision feedback and scoring. Revision memories can be shared and reused with new texts and with new trainees. The reporting module of the platform allows the profiling of translators (strengths and weaknesses) in an objective way