Google Translate

2022: Service outage in China

In October 2022, Google disabled Google Translate to users from China.

2017: Neural Networks

Google announced^[1] ​ ​ about a large-scale update of the Google Translate service: the work of the translator will be based on neural networks. For some languages, neural translation began to be used in November 2016; others, including Russian, will switch to the new technology in the near future. The developers claim that neural networks significantly improve the quality of translation, since machines can analyze not individual words and phrases, but completed sentences and the context^[2].

2016: Google's artificial intelligence invented its own language

The artificial intelligence of the Google Translate service, which began to use neural networks in its work when translating some language pairs, invented its own language, New Scientist reports, citing Google CNews employees^[3].

Google experts believe that the language that their artificial intelligence managed to create is an analogue of the international auxiliary language of interlingua.

Analogue of Interlingua

Developed by the International Auxiliary Language Association (IALA), Interlingua includes words with Latin roots as well as international vocabulary. Since the language does not use artificially invented word roots, as in many other artificial languages, but really existing Indo-European roots, interlingua can be understood by a huge number of people from different countries, especially those who speak any Romance language.

Similarly, as New Scientist writes, Google's artificial intelligence has created a common language for the convenience of translation, combining sentences and words similar in meaning and spelling from different languages.

At the same time, experts clarify that this new Google Translate language is still suitable exclusively for translation and is not suitable for communication between people. According to experts, the Google translation method, called Zero-shot Translation, is not able to compare with the "human" translation, but demonstrates good potential.

See also Artificial intelligence (AI) article Artificial intelligence (AI)

Google Neural Machine Translation

In September 2016, Google Translate began working on a new scheme using neural networks.

Now this scheme applies to eight language pairs (for example, English - Chinese) with which Google Translate works (in total, the translator supports more than 100 languages ​ ​ and about 10 thousand language pairs, translating about 140 billion words daily).

Details of the Neural Machine Translation (NMT) algorithm were published on the website of scientific papers arxiv.org. The authors of the work noted the difficulties of the method - not only the high cost of using NMT, but also problems with translating rare words.

2015

In January 2015, Google introduced an updated Google Translator app for iOS and Android. In the version of the service, it became possible to instantly translate inscriptions from English into Russian, French, German, Italian, Portuguese and Spanish, as well as from these languages ​ ​ into English. It is enough just to point the camera at the desired sign or text, and the translation will immediately appear on the screen even without an Internet connection.

"In the previous version of Google Translator, you could already use the camera to take photos of text and translate it into one of 36 languages. Now the translation will appear on the screen instantly. You will definitely not get lost in the tangled streets of Italy or you can easily choose your favorite dish from the menu in Barcelona Cafe! Simply point the camera at the desired sign or text, and the translation will immediately appear on the screen. It is especially convenient that this is possible even without an Internet connection, "Google said

2014

At the end of 2014, Google specialists are completing work on the simultaneous speech translation function in Google Translate. It is expected that this feature will be implemented in the version intended for installation on mobile devices built on the basis. operating system Android According to preliminary information, in the near future it will be possible to purchase a new product in a company store. Google Play It is planned that the mobile application will recognize speech in several languages, and the translation will be issued in the format of a text message. A similar service is already used in Google Translate, but the new version will offer an almost instant reaction, expanding the use of synchronized translation.

It is worth noting that a similar feature has already been implemented in Skype Translator from Microsoft. The service began to operate in testing mode in December 2014 for voice calls, offering real-time translation, but so far only operates in two languages ​ ​ (English and Spanish).

2012: App audience - 100 million users

The timing of reaching the number of 100 million users by various applications

2011

In April 2011, Google introduced an experimental voice recognition feature in the popular online translator Google Translate, the Google Operating System website reported.

The feature only works in test versions of the Google Chrome browser and can only recognize and translate phrases from English. However, since the speech recognition system is based on the developments of Google Voice Search, in the future the service will support more languages, including Russian.

The service uses the HTML Speech Input API system, which is responsible for voice recognition in Google Chrome. Support for this technology, introduced in Chrome in March, is still experimental. In the future, with its help, the developers intend to implement full voice control of the browser.

HTML Speech Input API is a relatively young extension of the HTML5 standard developed by Google. The specifications are not finished yet, so other browsers are not yet implementing support for this technology.

The technologies necessary for the full operation of the new Google Translate function are still at an early stage of development. For this reason, voice recognition often does not work in the right way. The developers have not yet announced when this feature will be suitable for everyday use.

See also Translate.ru from the company PROMT (PROMT)

Notes