Machine translation literacy and undergraduate students in applied languages: Report on an exploratory study

Rudy Loock,Sophie Léchauguette

doi:10.5565/rev/tradumatica.281

Abstract

This article reports on an exploratory study conducted on applied languages undergraduate students’ use of machine translation. Starting from the observation that they make extensive use of free tools available online, our aim was to understand whether they are capable of identifying and correcting machine translation errors, and if so, to what extent.

Highlights

With the ever-increasing place of machine translation (MT) in our everyday lives, especially since the advent of neural machine translation (NMT), students from all disciplines have been experimenting with the use of online, free MT tools such as Google Translate or DeepL for their homework assignments
Our aim was to evaluate our students’ MT literacy – precisely their capacity to deal with MT errors – and to check both the extent of their trust in the results provided by the machine, and the extent of their need for specific training to use MT tools effectively
Our results show that in general, students fail to identify most of the errors made by the MT tool

Summary

Introduction

With the ever-increasing place of machine translation (MT) in our everyday lives, especially since the advent of neural machine translation (NMT), students from all disciplines have been experimenting with the use of online, free MT tools such as Google Translate or DeepL for their homework assignments As a consequence, they need to develop a new competence, namely “MT literacy”, a concept put forward by Bowker & Buitrago Ciro (2019), for a “principled approach” to such tools (Loock, 2019). With NMT giving priority to fluency in the target language, sometimes at the expense of accuracy (see e.g. Bojar et al, 2016, Macken et al, 2019), errors are more difficult to identify, both for students and professionals (e.g. Castilho et al, 2017a/b, Yamada, 2019) This means that the notion of “trust” in MT output will be a key factor in users’ capacity to assess the appropriateness of what the machine provides. The question of trust boils down to the more general question of human-machine interaction: how to make sure that the human is assisted and not misled by the machine because of excessive trust in the information generated by online tools? As far as MT is concerned, this issue has become crucial for professionals and professionals-to-be in the translation industry, and for any user outside the industry because of the widespread availability of MT technology

Objectives

Results

Discussion

Conclusion