In the years when Google Translate was first available, it produced materials for internet comedy with some incorrect translations. Those who tried to talk to strangers with translation were frustrated by the other party’s inability to understand the sentences. In fact, the “chicken translate” joke, which many of us know, even took place in movies and TV series. Over the years, improvements in translations have started to be noticed in Google Translate, as in other translation applications. In fact, gradually the translation jokes began to decrease and the conflict between people due to mistranslation began to end.
So how did these visible improvements in Google Translate come about? How did quality translations that people chat with come about? Let’s examine it together.
Google Translate was working with statistics when it first came out.
Google Translate first appeared in 2006. In the first years of its release, it worked largely based on online statistics. When you typed a word, the system would scan the entire internet database, find available documents translated in the selected language pair, and display the most frequently used version of any phrase as a translation.
The initial working logic of the translation worked well for short sentences and single words, as the translation was easy to translate single words, but when it comes to languages with complex grammar and long texts, it has bad, funny results. In short, Google Translate could not produce meaningful translation results in context and sentence structure translations with its first technology.
Fundamental changes began in 2016 with “neural machine translation” technology.
Google said in a statement in 2016 that it had made a big improvement in web and mobile translation. This development was based on a technology called “neural machine translation”. This improvement first came to French, German, Spanish, Portuguese, Chinese, Japanese, Korean languages, which include Turkish as well as English. Translation application thanks to the technology called neural machine translation; it no longer divides the text into small chunks and translates, but instead translates whole sentences in phrases at once. This improvement results in better syntax and more natural and meaningful results.
In addition, Google Translate’s product leader, Barak Turovsky, said that this technological advance in translation is equivalent to 10 years of development. Fundamental improvements to Google’s translation app have come a long way in these years.
Offline translation has also improved by 20 percent in some languages.
The offline translation feature can be used with additional packages downloaded from within the Google Translate application. Online language packs are known for being used in places without internet access. This feature, which saves the lives of those who do not speak foreign languages and do not have internet, is perhaps one of the most popular and frequently used features of the translation application.
According to the statement from Google in 2019, Translation is available in 59 languages with the innovation of its offline feature; It provides 12 percent accuracy with word choice, grammar and sentence structure. This accuracy can also reach 20 percent in some languages such as Japanese, Korean, Thai, Polish and Hindi.
Translation added strength to its power by making use of artificial intelligence.
Google also uses AI (artificial intelligence) and algorithm-based learning methods. The translation app discovers millions of documents that have already been translated by humans, and based on these discoveries it can generate translations using patterns found in the text. Translation over time; it incorporates more patterns, receives input from real people, and improves its translations.
We see how the results improve when we put the puzzle pieces together.
In the 13 years since its launch, Google Translate has evolved with “neural machine translation, rewrite-based paradigms, techniques such as on-device rendering, and artificial intelligence.” Although these technologies developed by Google were not perfect in translation accuracy, they revealed visible changes compared to the past. At least we no longer encounter funny and strange results in translations.
In addition to these developments, Google Translate also has a volunteer community called the Google Translate Community, which consists of people who translate sentences and words, check whether the translations in the system are correct, and contribute to performance improvements in languages with low data. In addition to this community, an ordinary translation user can also contribute to Google Translate by scoring the correctness of their translations.
So what do you think about the translation results of Google Translate? Are you satisfied with the translations?
Sources: CNET, Google Blog, INC, TTC We Translate, Venture Beat
4845
