Translated by Charlotte Doane
Machine translation (MT) has undeniably made strides in the last few years. Take Google Translate, which used to provide results that were approximate at best and hilarious at worst and is now gradually finding a place in the world of language services and communications. While machine translation certainly has its advantages—mainly in terms of productivity and efficiency—it is also not without risk. And the truth is, it isn’t ready to replace human translators.
They see no reason to hand off the rewarding task of decoding a language, with all its subtleties, to a computer program.
Between Systran, DeepL, Tradooit and Google Translate, we have no shortage of platforms and software to choose from, and there are several that we use here at Cartier et Lelarge. While taking care to follow all confidentiality protocols, some of our translators regularly consult online tools to get ideas from the different combinations they propose, feeding the machine in the process. Others have installed Microsoft Word add-ins to improve their workflow. However, some of our team—though not for lack of skill or flexibility—adamantly abstain from using machine translation in their work. They see no reason to hand off the rewarding task of decoding a language, with all its subtleties, to a program that, in their view, was developed by IT specialists who did not see fit to consult any language professionals. And still others find that they already work so quickly, machine translation only slows them down.
Still, most of our colleagues say they use MT every day, though not for every project. Advertising and marketing content, which demands style and creativity, benefits much less from artificial intervention than general communications and certain administrative texts. Likewise, niche content is rarely improved by machine translation.
While machine translation can be good for troubleshooting and perhaps win us productivity points, we all agree it has a major downside: confidentiality.
As a general rule, MT is just one tool among many here at Cartier et Lelarge, which we reach for only when translation software, the Robert and Collins, Termium and Antidote come up short. It is particularly useful for time-sensitive jobs, in those moments at the end of the day when energy and concentration run low, or as a source of inspiration. One of our more senior colleagues likes to use it when revising to help her improve an awkward sentence here and there.
But while machine translation can be good for troubleshooting and perhaps win us productivity points, we all agree it has a major downside: it does not allow us to ensure confidentiality and the protection of our clients’ information. For this reason, we never enter any information that could potentially be used to identify a client into free online tools. This is crucial, especially given that machine translation companies can be located and store confidential data in different countries, potentially leading to problematic legal situations.
The products that have recently become available point to a bright future for MT technology.
We are also working on integrating an MT system into our pretranslation software, which suggests translations for segments that contain no terms from our translation memories. This approach has the advantage of incorporating our own body of work into the MT training for more precise terminology matches, but we have encountered problems due to the way our records are tailored to each client and domain.
For now, it seems that the next advances in MT technology will likely solve this operational challenge, and the products that have recently become available point to a bright future. The real challenge comes afterwards and will be up to translators to overcome. We know how to use machine translation to serve our clients, but how can we also serve our profession—without devaluing it in the process?
This article is part one of our series on machine translation. Read parts two and three.
