It’s no secret that India is one of the most linguistically diverse countries in the world. With 22 official languages and a hundred others that aren’t, breaking the language barrier can be quite a challenge for a business. But Google has made an effort to localize its products and services to a billion Indians and now shares everything it has achieved so far on its L10n India event.
Some of the announcements made at the event related to the availability of more languages in some of Google’s most popular apps and services. For starters, Maps in India is now available in 9 local languages, including Gujarati, Marathi and Tamil. Likewise, Google is making it easier to switch results to Indian languages - five to be exact – within Search.
An interesting tidbit from the event was that more people use Google Lens every month in India than anywhere else in the world. In order to better serve Indians, Google is now enabling the use of Lens – which can also help students solve math problems – in Hindi.
But the most far-reaching announcement at the event was that of MuRIL (Multilingual Representations for Indian Languages) – a new open-source multilingual AI / ML model that can use the knowledge, data, and learned knowledge of one language and in another. to apply. language. To address the idiosyncrasies of each language, Google would initially have had to build a separate model for each language. But given the enormous diversity, this would certainly not be the most resource-efficient way to break the language barrier.
With MuRIL, Google hopes to facilitate the development of Indian language technologies and deliver on its promise to bring the billionth Indian online. Since MuRIL is open-source, it is already available for download from TensorFlow Hub.