MT: Threat or opportunity? The example of post-editing

Machine translation (MT) engines like Google Translate used to be considered incapable of producing even a decent translation, let alone a good one. People thought that machines would never be able to replace humans for that task, because every translation is unique and a brain is needed to make the decisions, to choose which word to use. Still, a lot of progress have been made in that field (and I am not even talking about AI), with the use of corpus statistical and neural techniques. MT engines are not yet able to produce a decent translation of a long, complex text, but using it can give you an idea of the content of a text.

A concept closely linked to MT in the field of translation is post-editing (associated sometimes with pre-editing). It involves both machine and human translators, and is allegedly less time-consuming than classic, human translation. In post-editing, a human translator (the post-editor) corrects the translation produced instantly by the machine. The time required to review a translation depends on the level of quality expected. Light post-editing is less time-consuming but provides a lower level of quality, while full post-editing takes more time but gets a better result.

Another important thing to take into account with post-editing is the target audience. If the document you are translating is for internal use, then light post-editing of a raw MT output can be enough to make understandable or readable the document in the target language. On the other hand, if the document you are translating is for official use, then you will need a higher-quality translation, which will mean spending a lot of time on post-editing, which could turn out to be as or even more time-consuming than translating directly from scratch. The use of CAT (Computer-Assisted Translation) tools is going to be very helpful in spending less time on the translation, while ensuring the level of the translation, with tools such as translation memories and terminology databases.

Surprisingly, post-editing has been used as a way to improve efficiency by international organisations for more than 30 years. Nowadays, most translation agencies offer post-editing services. These services are cheaper than conventional translation services, but concerns only specific types of documents, and consequently they do not account for a big part of their work.

Post-editing has been widely criticized by professional translators, some saying that it was actually more time-consuming to make correction than to translate using CAT tools. Other pointed out the fact that post-editing was usually paid at lower rates than regular translation, and that therefore it was dragging down translation rates as a whole.

Still, technical progress alone cannot explain the increase in the use of post-editing. To understand this emergence, one has to look at how the world and companies have evolved during the last decades. With globalization, there are now more international exchanges than ever. A lot of multinational companies have been created, and international institutions play a more important role than before. All these changes have tremendously increased the demand in translation, making it impossible for human translators to take care of all of these workloads, hence the use of instantaneous, machine-conducted automatic translation.

Current computer-assisted translation tools now include MT tools, such as Google Translate, directly in their interface, paving the way to the use of post-editing, using both CAT tools like translation memories, and MT tools. More and more professional free-lance translators now undergo training to become able to use these tools and answer their client’s needs.

We can therefore say that MT is having a real impact on the profession, and translation professionals can no longer ignore its existence and the possibilities it offers. Using MT will mean, if not now, on the short term, a better productivity for translators while assessing high-quality translations. But won’t all these new tools change the very nature of the job, with translators gradually disappearing and being replaced by post-editing technicians? What is certain is that for the time being, humans will remain essentials to the production of a high-level translation.

Romain Simon