Google Translate: How does the multilingual interpreter really work? | The independent

Google Translate has become the go-to Internet resource for short and fast translations from foreign languages.

The service was first launched in April 2006, against early competition from Babel Fish. It now has more than 500 million users per day worldwide, offering 103 languages.

But how exactly does it work?

Originally focused on statistical machine translation, Google Translate worked by translating the required text first into English as the intermediate stage language and then into the target language, matching the sentence in question with millions of extracted documents. official transcripts from the United Nations and the European Parliament. .

Contrary to popular belief, Google has chosen not to hire experts to compile exhaustive grammars for each native language due to the constantly evolving nature of the language. Any attempt in this direction would quickly be rendered obsolete.

The original translations were inevitably flawed, but served to reveal the general intent of the original passage, even though Google Translate could not provide the smooth and flawless translation expected of a human expert.

But in November 2016, Google announced the transition to a neural machine translation premise – a “deep learning” practice that saw the service compare entire sentences at a time from a wider range of language sources. .

This ensured greater precision by giving the full context rather than just isolated sentence clauses.

By comparing the translations from Japanese to English with those from Korean to English, the service is able to infer and map the relationship between Japanese and Korean and perform round-trip translations between these two languages ​​in as a result, a big step forward in the understanding of computers. semantics, a process still confused by metaphorical expressions and eccentric idioms.

Processing these calculations over and over again allows Google to spot recurring patterns between words in different languages, which means its chances of achieving accuracy are constantly improving.

The result that use ultimately receives is finally refined to bring the passage closer to real speech.

Google Translate can be integrated into browsers like Chrome (where it has existed as a standard feature since February 2010), can speak highlighted text in many cases, and can spot words in images using recognition character optics and automatically identify unknown languages.

The company has offered an Android app since January 2010 and an iOS version since February 2011 and remains the first port of call for many looking for help with online copying in a foreign language.

[This article was originally published in June 2018]

Comments are closed.