You might have heard MTPE mentioned, perhaps you’ve seen it referenced on the internet. Those four letters have become quite commonplace in the translation industry, and they either induce a shudder, usually in us translators, or are embraced with open arms. MTPE stands for Machine Translation Post Editing.
Machine Translation is nothing new – Google Translate, for example, is a MT engine, a software application designed to automatically translate text from one language to another. PE is the stage that follows the MT, the editing of the automatically translated text by an experienced translator, ensuring that it sounds natural, adapting the terminology and making sure that specialist terms are appropriate to the content and audience.
MTPE offers a number of benefits, but drawbacks too. The most obvious benefit is reduced cost, but depending on the quality of the automatic translation, which itself depends on several factors, the amount of post editing required may well negate any cost benefit. While today’s MT engines are much improved when compared to the ‘junk’ that existed a decade ago, they still have a long way to go to match the quality and depth of expertise of an experienced human translator. For some texts they are better than others and for some purposes they are better than others – a text for basic internal communication purposes with fairly straightforward content can probably be machine translated, perhaps even without post-editing, without too much difficulty. It might sound awkward, unnatural and ‘clunky’, but it will do. Something intended for marketing purposes, to showcase a new product, set out detailed technical instructions or to communicate with an important client is probably best left to an experienced translator.
The idea of machine translation makes many professional translators shudder, while some feel that it makes them more productive. I can’t deny being somewhat wary of it, as well as slightly irritated. After all, most translators spend years of their lives working on their education and training and amass years of experience before they call themselves ‘professional’, so to see a server in a data centre somewhere do what you’ve been working towards for years in just a matter of seconds can be pretty disheartening. Still, my own experiences with it have showed me that it’s not all that. Estimates are that in terms of time, it saves a translator anywhere between 10% and 60%, up to 100% in some case. My trials on a few documents of varying type and length indicate that yes, for some very simple sentences, 100% is possible, but that 10% is probably more likely. Syntax I found to be a problem, specialist terminology and consistency