Bhashini, India’s AI-led language translation and database platform, will soon start charging enterprises and institutions based on features and service requirements, CEO Amitabh Nag told FE.
“We will be charging for many services which will come up over a period of time. We have already received orders worth crores from our customers. Government as well as private organisations have shown interest,” Nag said, adding that Bhashini is also doing proof of concepts for multilingual call centres or interactive voice recognition (IVR) setups.
As Bhashini is working with different companies and government institutions with regard to specific use cases, over the next 2-3 months, it is expected to offer its real-time language translation service on a paid basis. Prime Minister Narendra Modi also recently used the platform to translate his speech from Hindi to Tamil on a real-time basis.
“We are launching an application which will give the demo (for real-time translation) in open format, and after that, we will be launching that as an offering on a paid basis,” Nag said.
Bhashini, developed by the Digital India Bhashini division under the Digital India Corporation, a Section 8 company, under the electronics and IT ministry, will complete two years in August.
Since its launch, it has enabled features like text-to-text translation in 22 languages, automated speech recognition, text-to-speech synthesis, optical character recognition, video translation, document translation, language detection, voice-based payments, among others.
Currently, there is a demand for live speech translation in various sectors. Nag explained that the service can be used for domains like judiciary, education and skilling, which can be multilingual with real-time translation. Bhashini’s real-time translation works in 12 languages and it has improved the live translation lag from 6 seconds to 1-2 seconds now.
Asked about the use of Bhashini’s features by political parties for poll campaigns, Nag said, “we are not part of any political campaign, we are part of government programme. What we do specifically is based on specific requests, and that is something decided by the government”.
Currently, 40 million inferences per month are happening on the platform. Bhashini has an app which can be used by people. It also provides API (application programming interface) integration to startups which want to use Bhashini’s language features on their platforms such as chatbots, etc.
Inferences mean translations happening on the platform by users using its different features. The National Payment Corporation of India (NPCI) has collaborated with Bhashini to allow users make conversational payments through UPI in Indian languages.
“We are now able to deal with multiple customers who are using our NHLT (national hub for language technology) platform. Some of them are using it on a proof of concept mode, some are using it on an end-user deployment mode, and the deployment is increasing,” Nag said.
Besides fintech, Bhashini is also working on solving problem statements for the e-commerce sector, which it received from the Open Network for Digital Commerce (ONDC). “There are multiple use cases for e-commerce. The NHLT platform itself can become multilingual and voice-enabled. So, you can build the cart using voice. The inventory management can be multilingual as well as voice-based,” he said.
At the core of Bhashini lies the collection of datasets in regional languages. For this, the organisation follows a crowdsourcing model called Bhashadaan, where people can contribute to these datasets by speaking, validating, writing, etc. “I believe the traction on the platform (Bhashadaan) would now increase because of people wanting to contribute. The platform is getting built up for more robust systems,” Nag said.
Bhashini is working with over 70 research institutes for development of AI models. It will be growing to about 100 people in the next 3-4 months, and in addition, has a contractual strength of another 100. It also has about 200 translators working on the field to gather data.
“The overall objective is inclusivity and empowerment. So, whatever is required to be done, from a language perspective, we will be doing. We have to look at sectors which are uncovered,” he said.
