Recently, Google Translate has quietly upgraded its translation kernel. Google’s neural machine translation system, gnMT: Google Neural Machine Translation, uses the most advanced training technology available, improving machine translation and reducing translation errors by 55-85% again, according to google official data.
The quality of the translation model sited by Google
More than a decade ago, Google released Google Translate, an early phrase-based statistical machine that breaks down input sentences into words and phrases and translates them independently. The disadvantage of this translation is obvious: the original complete information in the sentence is fragmented and cannot be expressed in a coherent manner. This phenomenon is particularly evident in the case of English-Chinese translation.
Google Neural Machine Translation, on the other hand, translates the entered sentences as a whole.
In the case of Chinese-English translation, Google Neural Machine Translation first encodes the word Chinese into a list of vectors, each of which represents the meaning of all the words read so far (encoder “Encoder”). Reading the full sentence, the decoder starts working — one word at a time for generating an English sentence (decoder “Decoder”).
The image above shows the principles of Chinese-English translation of Google’s neural machine
In order to generate the correct words at each step, the decoder focuses on the weight distribution of the Chinese vectors that are most relevant to the encoding of the English word generation.
When first proposed, the neural machine translation system was comparable to the phrase-based translation system on a medium-sized data set.
Now, for its part, Google says it has built a system that translates better at speed and accuracy by enabling neural machine translation to overcome many of the challenges of working on very large data sets.
At present, Google’s neural machine translation system has been invested in Chinese-English translation. Now, both mobile and web-based Google Translations are fully translated using nerve machines – about 18 million translations a day.
With a large data set tested by Google’s neural machine translation system, what about the latest Google translation?
We did a simple comparison test. Of course, it’s still in the Chinese-British translation scene.
Test Scenario: PC-side Google Translate web version
A random selection of an English message, the original text is as follows:
Since COVID-19 began, we’ve’ve heard from ours retail and brand manufacturing partners at its’re hungry for more insights on how consumer interests are changing, we’ve heard from ours demand. We see these changes reflected in how people are searching on Google. Last month, there was ed spikes in search interest for household supplies and jigsaw puzzles as people sss more time at home. This month we’ve seen surging interest for sewing machines and baking materials in the U.S., and tetherball sets and chalk in the United Kingdom and Australia.
Businesses are using a variety of resources to understand changing consumer interests—including Google Trends, social listening, surveys, and their own data—in order to help make decisions on The fly. But if they don’t know what to look for, there isn’t an easy way to understand what’s the product categoris are gaining in canity, and ed pose an OPPOrtunity.
That’s why we’re launching a’s rising sares tool on Think with Google. It Surfaces fast-growing, product-related categories in Google Search, the locations where they’re growing, and the queries with with. This is the first time we’ve ve provided this type of insight on the product categories people are searching for.
The results of the English translation given by the old Google Translation:
After the upgrade, Google Translate gives new results in English translation, and the red mark word section is different from the old translation results. The translation of the new version is as follows:
Comparing the results of the two translations, we can see that the difference is quite large. On the whole, the expression of the red text in the new translation results is obviously more in line with the Chinese grammar and expression habits.
In addition, the optimization of the translation results in the last paragraph is more obvious, the Google product name “Think with Google” was accurately identified, avoiding the embarrassment of mistranslation.
It’s really smarter!