What is the Multilingual Representations for Indian Languages (MuRIL) launched by Google at its L10n event?
Google unveiled a new AI-powered language model 'MuRIL' on 17 December 2020 at its L10n event. Users can now toggle search results between English and any of the four newly added Indian languages-- Tamil, Telugu, Bangla, and Marathi-- on mobile devices. Also, Google Maps (navigation app of Google) will now support nine Indian languages.
.@partha_p_t, Research Scientist, Google India, talks about MuRIL - a powerful multilingual model which is free, open-source, and available for all on the TensorFlow Hub 👨🏻💻👩🏻💻— Google India (@GoogleIndia) December 17, 2020
Hear about this on the #L10n livestream ➡️ https://t.co/wHMPTokz8C pic.twitter.com/I4Nyx0ZRoa
In July 2020, Google made an announcement to accelerate India’s digital economy by investing $10 billion in India over the next 5-7 years through its 'Google for India Digitization Fund'.
The investment aimed at four key areas:
1- To enable affordable access and information for every Indian in the language they are familiar with-- Hindi, Tamil, Punjabi, Telugu, Kannada, etc.
2- To build products that fulfil the unique needs of the Indians.
3- To empower business as Indians continue to make efforts towards the digital transformation in the country.
4- To leverage technology and AI for social good, in areas like health, education, and agriculture.
What is MuRIL?
Multilingual Representations for Indian Languages (MuRIL) is the latest multilingual model launched by Google and is aimed at improving interoperability from one language to another. As per Google, MuRIL supports 16 Indian languages.
Features of MuRIL:
Let us have a look at the wide range of features which will enhance language experience for millions of people pan India:
1- Easy toggling between English and Indian language results.
Users can now toggle search results between English and four additional Indian languages-- Tamil, Telugu, Bangla and Marathi-- on mobile devices by using 'chip' or 'tab' below the search bar.
2- Understanding which language content to surface when the query is searched in English.
As per Google, typing in an Indian language in its native script is difficult. As a result, many people search in English but prefer the search results in their local language.
Thus, in the coming months, the search will show relevant content in supported Indian languages where appropriate, even if the local language query is typed in English. It will roll out in five Indian languages-- Hindi, Bangla, Marathi, Tamil, and Telugu.
3- Enabling people to use apps in the language of their choice.
As per Google, more than 50% of the content viewed on Google Discover in India is in Indian Languages. Also, after the launch of Assistant language picker for Google Assistant, queries in Indian languages have doubled.
Now, users across India can use one of the nine Indian languages on Google Maps by tapping 'App language' in Settings. This will allow the users to search for places, get directions and navigation, and interact with the Map in their preferred local language.
4- Homework help in Hindi and English.
Google Lens helps you search what you see, get things done faster, and understand the world around you—using just your camera or a photo. Now, users can use Lens to snap a photo of a math query and learn how to solve it on their own, in Hindi (or English).
5- Helping computer systems understand Indian languages at scale.
With the launch of Multilingual Representations for Indian Languages (MuRIL), the transliterated text is now supported-- a feature missing from previous models of its kind.
MuRIL is also good at determining the sentiment of the sentence. For instance, "Acchha hua account bandh nahi hua" would have been interpreted as a negative statement in the past but with the launch of MuRIL, the statement will now be identified correctly as a positive statement.
At present, MuRIL supports English along with 16 Indian languages and can be downloaded from the TensorFlow Hub, for free. It will help researchers, students, startups and anyone who is interested in building Indian language technologies.
In October 2020, Google added a new clause in its publisher policies to stop monetisation of new content, web pages and applications in regional languages which are not supported by Google and will come in force from March 2021.