Google Translate is now aiming for 1,000 languages


In recent years, thanks to artificial intelligence, machine translation has continued to evolve. And today, thanks to applications such as Google Translate, we have the possibility of consulting information in Mandarin, Hindi or Japanese, without understanding these languages.

In addition, Google Translate is no longer content to translate text files. The app also allows us to translate writings on the camera or on images, and when we travel, it can also serve as an interpreter.

But although Google Translate already allows us to accomplish a lot of things, in reality, Google’s machine translation app is still very limited, considering the number of languages ​​there are on the planet. Indeed, it, like the web in general, has focused on the most spoken languages ​​in the world.

Google announces a new initiative for its translation

The good news is that Google is about to fix this. This year, the firm added 24 new languages ​​to its machine translation, including Bambara, Quechua, and Sanskrit.

However, despite the regular addition of new languages ​​to the app, Google Translate only supports a very small portion of the more than 7,000 languages ​​spoken around the world. And Google’s new challenge is to support 1,000 within a few years.

To achieve this, the firm is launching a new project called “1,000 Languages ​​Initiative”. Based on artificial intelligence, the project aims to offer automatic translation for the 1,000 most used languages.

“It will be a multi-year undertaking […] but we are already making significant progress here and seeing the way clearly. Technology has evolved at a rapid pace – from how people use it to what it is capable of”said Jeff Dean, senior vice president at Google Research.

In order to achieve this objective, and bring “greater inclusion for billions of people in marginalized communities around the world”Google intends to use all formats, not just text.

Indeed, in the world, people share information via text, but also via images, speeches and videos. However, the firm’s models are now “multimodal”, which means that they are able to extract information from these different formats, in order to translate new languages.

A new AI, and data to exploit

Google has also already developed technology that exploits these new opportunities. “As part of this initiative and our focus on multimodality, we have developed a Universal Speech Model – or USM – which is trained on over 400 languages, making it the largest linguistic coverage seen in a speech model. word to date”says Jeff Dean.

Besides that, Google will also develop its translation working with local communities, in order to get more data for its AI. In Africa, Google is working with organizations and researchers to bring Gboard voice input to 9 new African languages. And in Asia, the firm even works with governments.

Leave A Reply

Your email address will not be published.

rabit tradon seo tool seo hyperlink with keyword generate reset