Google translate

How does Google translate work?

If we affirm that we have all used the Google translator, we are not generalizing. The platform lives in a process of continuous improvement and this has been demonstrated in recent years. It is an indispensable tool that has gotten us out of trouble, but how does this technology that simplifies our lives work? Google Translate currently works with 103 different languages ​​and is used by more than 200 million people every day. And, despite the fact that it was released in 2007 and has been with us for more than ten years, its operation has surprised us. Neither linguists nor philologists, we reveal what is behind the scenes of Google Translate.

✓Use Google translate from any app in Android - YouTube

First step: dive into translated texts

Managing 100 languages ​​must be a really exhausting task for Google programmers. Well, it is easier than it seems a priori if the translation is achieved through automatisms . Google Translate works with  machine learning through examples .  This is how Peter Norvig, Director of Research at Google, explains it in  one of the company’s blogs . At the beginning of the tool, the programmers made an effort to ensure that the technology at hand understood every word. This meant, and never better said, that he was able to differentiate between grammatical categories, understanding the complexity of each language . What happened is that they found themselves with a job that was impossible to materialize. Millions of words, billions of combinations and different rules for each language. The solution to deal with such a large amount of information was to prepare the translator to immerse himself in the web. The programmers thought that the best way to find patterns and examples was to locate them in the millions of already translated texts available on the web. In this way the translator could look at complete sentences and take context into account, not treat each word individually.

The software that changed everything with basic English

The automatic functionalities provided by Google are achieved through software that performs “statistical machine translation” (SAT) . As they say in websa100 , what the assistant does is search for similar patterns by diving through millions of documents. These are widening daily and improve our experience.

The software that changed everything with basic English

The automatic functionalities provided by Google are achieved through software that performs “statistical machine translation” (SAT) . As they say in websa100 , what the assistant does is search for similar patterns by diving through millions of documents. These are widening daily and improve our experience. The gear that makes everything develop correctly is English . The TAE  “works by first translating into English as an intermediate language of passage and then into the target language, cross-referencing the sentence in question . ” That is, instead of directly translating the language, the text or word is first translated into English to return it in the desired language. This operation greatly optimizes the work of the translator. English works as a link for a compelling reason . The availability of texts translated into this language are countless. Millions of documents that the platform has saved for itself. What allows a translation to be accurate, improvable, or downright horrendous, is the material available to Google . English is the ideal pass because translations are likely to be available in almost any language. The number of translations between English and Telugu (the language spoken in some Indian states), for example, will be greater than between Telugu and Spanish. And so with every language used in the tool. Statistically the translation will be much better if it is first translated into English.

Google’s competition, its greatest ally

At this point we can say that the processing of grammar achieved by Google Translate is genius. And improving the user experience can become a pitched battle for the California company. From Inverse they warn of the strong competition of companies to develop the most efficient technology. The obvious thing turns out to be that the place of origin of the company determines how good they are in their field , and thus they get Google to also keep an eye on them to improve their translation in that language. Why? Because languages ​​with many digitized and quality texts will have a better translation than those that do not have them. If a language lacks texts translated into English, it is likely that the result will limp (and it is in these cases when Google requires external help). The bigger the database, the better the grammar will be processed and the more precision the user of the tool will achieve.

Leave a Reply