Google announced on November 2016 that its language machine translation service, commonly known as Google Translate, is approaching human-level accuracy. In fact, the search engine giant said that the new technique used, called the Google Neural Machine Translation system, or GNMT, reduces translation errors by an average of 60% compared to the previous phrase-based statistical translation system.
Several tools, including MateCat, leverage on Google Translate APIs to provide translation suggestions to translators. So this is an important milestone for MateCat too, considering that we are going to integrate Google Neural Machine Translation APIs as soon as they will be available for developers.
The MateCat team started testing the new algorithm a few minutes after the announcement. We are all very curious to measure the results and see if they match with the (optimistic?) figures announced by Mountain View.
Google launched Google Translate ten years ago with the goal of breaking language barriers and making the world more accessible. So this is a very exciting news for the 500 million Google Translate users community worldwide, even if the GNMT has not been yet deployed live for all language combinations, except for the Chinese to English language pair.
And this is very good news for MateCat users too:
- MateCat has partnered with Google since the beginning, leveraging Google Translate’s APIs and basing its value proposition on technology and post-editing with the aim of making the translation process easier, faster and more efficient;
- We see that machine learning is making major progress towards accuracy, and there is a clear trend of improvement in quality in the next few years. Translators and clients will both highly benefit from the upcoming revolution: human-quality translations will become increasingly accessible to everyone;
- We believe that technology and innovation are key in the industry. Our philosophy is to have machines and humans working together: humans are more creative and smarter than machine and they shouldn’t be bothered with repetitive and silly tasks. They are more suited to machines.The rest, the creative part of translation, will still be in the hand of professional translators.
Google will be rolling out its new method to the rest of language pairs in the coming months. We’ll make sure that the MateCat users will be the first to benefit from this breakthrough!