Are you paying too much for translation?
Dramatic improvements in machine translation (known as “MT”) are coming at us so fast that the translation market has hardly had time to adapt. As a result, current market pricing for translations has not yet caught up with this technology’s ability to lower prices.
It’s important to note that this article does not claim that MT can replace humans; that’s the stuff of science fiction. But it does claim that MT should already be reducing market prices in an increasing number of language combinations, subject matters, and use cases.
So where exactly do the cost savings come from? MT is now so good for general content in a lot of popular language combinations that a human can fix a machine translation in half the time it would have taken them to translate it from scratch, if not faster. This is called “Post-Edited Machine Translation”, or PEMT, and in a fast-growing number of scenarios it’s as good as, yet much cheaper than, an old-style translation. With PEMT, a computer translates the entire text as a starting point that can be surprisingly good depending on the languages and subject matter. A skilled professional translator then edits each machine-translated sentence, correcting errors and improving readability. PEMT can employ heavy editing, resulting in great translations, or light editing, resulting in “good enough” translations. The price is, of course, proportional to editing time, but it’s always less than a traditional translation. As the quality of MT continues to trend upwards, translation prices will (or should) trend downwards. Quality, of course, will remain constant.
So why haven’t prices gone down already? Because the translation industry is in a chaotic gray area, somewhere between not being convinced of the merits of MT and not willing to change their legacy systems and processes. While some leading-edge providers have embraced MT and are seeing good results, most have only experimented with it and remain unconvinced, continuing to work under traditional models with traditional pricing. It takes a dedicated MT strategy, rather than just ad-hoc experimentation, to understand when and how to benefit from MT. It’s also likely that some translation providers may be taking advantage of the chaos by using MT to reduce costs but not passing on the savings because the market has not yet demanded it.
Is your provider maximizing the use of MT and PEMT for appropriate scenarios? Ask them. If they are, are you seeing your per-word rates trending downward? If the answer to either question is “no”, then perhaps you’re not yet benefiting from recent advances in MT technology. You’re pricing is under the “old” model where a human translator translates every sentence and perhaps a second translator reviews. Glossaries and historical translation memories help speed things up, but a human is still translating each sentence.
Is all of this bad news for the translators themselves? Not at all, but it does signify a change in how they will work. While there will always be a demand for pure translation, for example in creative marketing and literature, the mainstream work of translators will trend towards editing. Translators who can adjust to this change will continue to thrive. They will be paid less per word, but they’ll process more words per hour. In fact, given that the amount of translatable content out there increases significantly each year, as translation prices drop, more content owners will see the ROI in translating. This should create even more demand than today. If you’re a translator, now is the time to learn to edit machine translation and adjust your pricing appropriately.
The future is here, folks, and it’s bringing your translation rates down. Or should be…