Google Translate switches to Neural Machine Translations
Google announced yesterday that it switched the algorithm that powers its Google Translate service to Neural Machine Translations for eight language pairs, promising that more will follow over time.
Google Translate is a useful translation service that is built-in to Google Chrome but also available as a web service and as mobile apps. This first step into the future of Google Translate enables the functionality for the mobile apps and Web service. What's not yet include are web site translations but that is to follow eventually as well.
Neural Machine Translation has been enabled for the following languages: English, French, German, Spanish, Portuguese, Chinese, Japanese, Korean and Turkish. These languages cover 35% of all Google Translate queries according to Google.
Today we’re putting Neural Machine Translation into action with a total of eight language pairs to and from English and French, German, Spanish, Portuguese, Chinese, Japanese, Korean and Turkish. These represent the native languages of around one-third of the world's population, covering more than 35% of all Google Translate queries!
We talked about the switch to neural networks before in regards to Google Translate, and suggest you check out the initial article that provides details on how neural networks differ from traditional computer powered translations.
Only this much: One of the core differences is that neural machine translation looks at the sentence as a whole instead of just individual words or phrases. This improves the translation significantly according to Google as context is better understood which in turn produces better translated sentences.
At a high level, the Neural system translates whole sentences at a time, rather than just piece by piece. It uses this broader context to help it figure out the most relevant translation, which it then rearranges and adjusts to be more like a human speaking with proper grammar. Since it’s easier to understand each sentence, translated paragraphs and articles are a lot smoother and easier to read. And this is all possible because of end-to-end learning system built on Neural Machine Translation, which basically means that the system learns over time to create better, more natural translations.
While your mileage may vary, neural network translation improves the quality of the translated text significantly when compared to Google Translate's previously used translation algorithm.
Companies with access to Google's Cloud Translation API may use the new translation system as of today as well.
Closing Words
Google Translate was always one of the better translation services on the Internet. While far from perfect, it usually produced better or at least similar results than comparable services. The switch to Neural Machine Translations will improve translations significantly. While limited to eight languages currently only, the new system will be available for all 103 languages that Google Translate supports eventually.
Now You: What's your take on the quality of the "new" translation algorithm?
Hello Martin,
First of all, thank you for your great work, I’m a long fan and reader of gHacks! Ive got a little problem with the site although (not related with this post). I can’t see any comments on mobile (clean browser, no Adblock etc.) Is it something wrong on my side? Does anyone else experiencing similar issue either?
Hi Duri, we are testing a new theme currently and it does not include comments right now. We are aware of the issue though and are exploring various options to change that. Sorry for the inconvenience.
I use Google Translate to translate a substantial amount of text from time to time which spares me a lot of time and work of having to do so myself manually, I have to proof read the translated text for correctness once the text has been translated in bulk which is where the translation work begins on my end, the implementation of this neural machine system will allow for more accurate translations to be made, obviating the need for me to make further corrections
I tested Google Translate just now, I can confirm that many of the common prior errors that I would be output in translations are no longer being made
United Nations translators aren’t out of a job yet, but this step is a major useful one to me, as I’m stuck with English only, with my personal lack of linguistic skill. I long ago forgot the little German, French, and Italian I used to know.
Hoping it improves radically. It’s Japanese English has been horrible. Japanese has three different alphabets, so it can be a pain to figure things out. Especially since they mix all three into sentences.
If you are a complete dummy this change may help.
But if you have even a small amount of knowledge this change is a disaster at least English Spanish. You no longer have the option of a looking at alternate translations of individual words or phrases that are wrong for your situation or use or country. Then change options offered to the correct result. This happens most of the time..
The only solution is to break up the sentence into vertical individual words or phrases to edit for the correct result.
Couldn’t agree more. I use bullet points a lot of the time and it just doesn’t seem to get the “sense” of the sentence. i much preferred being able to highlight an individual word and then being able to alter it to suit the sense.
I hope this update helps to put an end to Chinese to English translation fails that are often observed on signs
tinyurl. com/ksv6abu
it has to stop(!)
I have been using Google translate for four years in my bilingual work. Now without a drop down box that provides options in translating, I am forced to switch to some other translating service – NOT Google.
My friends and I agree wholeheartedly with the above assessment: “Google Translate was always one of the better translation services on the Internet.” “Was” is the key word.
EXACTLY! It may be better for a village idiot who knows nothing about the target language.
But if you just need help it is useless. except for one “trick” which is a PIA.
Separate the sentence into words or phrases with carriage returns (“Enter” key) then each can be edited.
Then after the past put them all back together in sentences (“Backspace”) key where needed.