So, what exactly is NMT?
NMT (Neural Machine Translation) is the latest trend in automated translation. Neural algorithms are the foundation of this new technology. NMT uses neural networks to automatically translate source texts into other languages, without human intervention. The system can then be improved (machine learning) to achieve better results over time. The results obtained clearly surpass those offered by previous technologies (RBMT or rule-based machine translation, and SMT or statistical machine translation) and some translations are close to professional translation for certain language pairs, including French to English, for example.
In this model, sentences are considered as a single unit; this differs from other MT models where the unit is the word. Result? Significant improvement in grammar, style, fluency and consistency. Based on machine learning, the neural translation engine can learn continuously and thus further improve the quality of the translations produced.
And what about the final quality?
Since Google Translate’s first beginnings in 2006, automated translation has had rather bad press, and with good reason. Old rules-based systems or statistical models offered disappointing results. However, with the acceleration of progress in machine learning and artificial intelligence, the new place of MT is that of a translator’s tool, just like the CAT tools or grammar checkers. One thing however is sure: the translator has the last “word” on the final translation. This is called the post-editing process, in which the expert linguist reviews and corrects each sentence. Integrated quality controls and spell checkers, together with this careful revision, guarantee the excellent quality of translations produced through NMTPE (neural machine translation post-editing), in all aspects equivalent to human-only translation.
What’s in for me as a customer?
With the help of artificial intelligence, (technologically advanced) translation companies that seize the opportunity can now easily offer same-day and next-day delivery options. Where the translation of a 15-page document would take half a week to a translator, it can now be done in less than one day, review, proofreading and layout included. Or in a few hours, in the case of a small team of translators. That’s a major leap ahead!
This increased productivity also frees up time for other tasks, such as better planning your linguistic and brand identity. Your lexicons and your glossaries can be improved and detailed style guides can be created to guide the translators in their new postediting roles. Or why not take some of that time to translate more content, this thick products catalogue that your customers keep asking in their language.
Then, should I expect to pay less?
The various IT investments required to implement advanced machine translation solutions and other cloud-based computer-assisted translation tools are significant. Some of them are license-based and must be paid every year while others are on a per-usage basis. The continued improvement of custom NMT systems requires specialized knowledge and hiring personnel with new skills. Translation agencies, like any other businesses, also have fixed costs such as office space, computers, electricity bills and other fees that can’t be optimized. One must also consider the cost involved in training the employees transitioning in their new roles, because current University programs are lagging behind in terms of post-editing training.
Short term, you should expect prices and quality to stay the same and keep an eye on those delivery speeds, that will continue improving. We are not on the verge of a major disruption of the language professions anymore. It is happening right now. For the good and the bad; hopefully mostly good.
Now, how about this….This article was postedited from one language into the other using a state-of-the-art neural translation engine. Hard to tell which one is the original, right? We agree.
NMT, SMT, postedition….so you learned a lot of new trendy words today! If your translation provider does not talk the talk, come see us online. We’ll be happy to assist you.
Deep Blue was a chess-playing computer developed by IBM. Deep Blue won its first game against a world champion on 10 February 1996, when it defeated Garry Kasparov in game one of a six-game match.