India is a multilingual nation. Over 300 million Indians use smartphones, however, there is a significant number of users which are still wary of adopting the web world owing to the gap between how the services are offered and the way they are consumed. Having a smartphone is a boon in the digital age, but is the language becoming a barrier for the majority of Indians from tapping the fullest potential of a smart device or internet in general?
Internet giant Google sees an opportunity of growth in the vernacular segment. While it has already added Indian language support to some of its services, the company today announced further expansion to the number of Indian languages supported. It also revealed plans to leverage machine learning to further improve its services with the Indian languages. Starting today, Google‘s products including Maps, Translate, Chrome, and Gboard will support over 30 Indian languages.
To help make web more inclusive for the Indian users, Google conducted a study in collaboration with KPMG. As of now, there are 234 million Indian language internet users as compared to 175 million English internet users. It is estimated that by 2021, Hindi speaking users will overtake English speaking Internet users. Furthermore, 9 out of 10 users in the next four years are likely to be Indian language users.
With the tremendous opportunity to make Google a more household name, the company already has a slew of products catering the Indian users. From Internet Saathi, GStation to YouTube Go; Google has lanched a slew of services and products customized for the Indian users.
It has been found that about 60 percent of users believe the lack of support for Indian language keeps them away from exploring the buzzing world of internet. Hence, improving on its support for Indian languages, Google also extended its neural machine translation system for as many as nine Indian languages.
Using the neural machine translation system, Google tools such as Gboard, Maps, Chrome are able to better translate the input as per user’s requirement. As Melvin Johnson, Google engineer, puts it, “…over 90 percent of Google Translate traffic comes from outside the US and India is seen as one of the largest and fastest growing Google Translate countries.”
At the heart of the neural machine translation system is a multilingual model which helps in scaling up the system. It basically allows one to put an input in local language and get results in a different language seamlessly. So for an example, the system can do translations from English- Hindi and English- Spanish; so the improved system basically suggests a possible translation from Spanish to Hindi or vice versa through learning.
The time involved in training these language models requires between two to three weeks per model. Google has also been crowdsourcing ‘human quality’ translations. The translation community was launched first in 2014 and addresses the challenges of translation for 44 languages.
With these systems at the core, Google’s multilingual products like Gboard, Chrome, Maps, etc can now support translations from Hindi to English, or English to Tamil in a better fashion. Going forward, Google aims to not only to bring more support for the Indian languages but also build an ecosystem where Indian language users are able to chat, shop, advertise, pay, search, and connect in the language of their choice.
0 comments: