Post-editing vs. translating in the legal context: Quality and time effects from English to Spanish

Abstract

Limited research has addressed the use of machine translation (MT) with legal texts, while recent data-driven MT approaches have improved quality. The present study reports on the results of an experiment involving 26 translators who post-edited and translated legal texts from English to Spanish. The results show that post-editing resulted in quality and time gains, with the former being statistically significant. Quality was determined by using standardized scoring criteria from a professional association, and the time taken to complete tasks was logged using specialized software. Participants completed questionnaires before and after the experiment to gather data about demographic variables, prior experience, and perceptions of MT post-editing, and their reflections on the experiment. There was general agreement among the participants that the machine translation output was beneficial, namely at the terminological and phraseological levels. However, participant variables related to years of experience or translation credentials did not reveal a demonstrable impact on quality and time.

Keywords

• machine translation
• neural machine translation
• legal translation
• post-editing