How to get AI to work in its 22 languages

Vineet Sawant has spent the past two years moving in the streets of Mumbai on a motorcycle as a delivery driver.

“Be on the road is always very stressful, especially in cities like Mumbai,” he says.

But when the linguistic barriers began an additional problem.

His first language is the marathi, and the English Mantha Sand is very little. “I can understand but very difficult to read,” he explains.

That caused problems in his new job.

He said: “Initially, it was difficult. Everything was in English, and I could understand some of it, but I am more comfortable in the marathi. I used to ask other delivery youth to help me know what to do.”

The employer, Zepto, is “the fastest delivery of grocery online in India.” So the presence of drivers who struggle with delivery instructions was not perfect.

To soften this process a year ago, Zepto has made a partnership with Reverie Language Technologies to provide Amnesty International’s translation service to its drivers.

Since then, delivery drivers have been able to choose between six languages on the Zepto app.

“I no longer have to guess,” says Sawant.

“Earlier, I will take more time to read and sometimes I made mistakes. Now if the customer writes” Ringing Bell “, I get these instructions in the marathi. So, I shouldn’t ask or check again. Everything is clear.”

The difficulties of Mr. Swanant are common.

“India has 22 official languages and hundreds of dialects,” says Professor Bohbac Bhazchariya, from IIT MOMBAI, a senior expert in India in using AI in Indian.

He points to “without technology, and this understands and talks about these languages. Millions of digital revolution – especially in education, governance, health care and banking services are excluded.”

The display of new artificial intelligence systems, such as ChatGPT, makes the task more urgent.

Huge amounts of data, such as web pages, books, or video texts, are used to train artificial intelligence.

In widely spoken languages such as Indian and English it is relatively easy to get, but for others, it is more difficult.

“The main challenge to create Indian language models is the availability of data. I am talking about repeated data. Coarse quality data is available. But these data are not very high quality, they need to be filtered,” says Professor Bhatersia.

“The issue in India is for many Indian languages, especially tribal and regional dialects, this data is simply not present or non -digital.”

Reverie Language Technologies is now publishing AI’s translation technology for a group of Indian companies.

The co -founder Vivekananda Pani says that although translation technology will make communication easier, there is “the possibility to push less common accents aside.”

“The challenge will be to ensure that the amazing benefits of linguistic developments driven by artificial intelligence are not mistakenly rich in human language.”

To help address the problem of Professor Bahtasharia He contributed to BhashiniA government project to develop high -quality data sets needed to train artificial intelligence.

In addition to data collections, Bhashini has built AI language models and translation services in 22 languages.

It started in 2022, it is a huge task, but it has already made a lot of progress.

Bhashini currently hosts 350 linguistic models that depend on artificial intelligence who have tackled more than a billion tasks.

More than 50 government departments work with Bhashini, as well as 25 government government.

For example, Bhashini Tech is used in multi -language chat groups for public services and translated government plans into local languages.

“Bhashini guarantees the linguistic and cultural representation of India by building the models of Amnesty International for India instead of relying on global platforms,” says Amitab Nag, CEO of Digital India.

It is hoped that the rural users will have within the next two or three years to reach a full sound to the government services, financial tools and information systems in their original languages.

We hope that these day -of -India -based data collections will give people who develop artificial intelligence models tools to make it easier to adapt to the whole population.

Currently, the design of any Amnesty International program to deal with complex operations such as health care can be very difficult.

KSHITIG JADHAV, Associate Professor at the Koita Digital Health Center at IIT MOMBAI, is working on Amnesty International that will help people stop smoking.

He explains that people in different stages of the process need different advice and usually need a well -trained person to conduct this evaluation.

But there are a limited number of practitioners who can help, especially those who can work in multiple languages, so Professor Jadhav hopes that the artificial intelligence model can bridge the gap.

Professor Jadaf says that artificial intelligence “will first define the type of conversation that a person needs, and accordingly, he will postpone questions and show sympathy and emotions.”

And all of this, we hope it will eventually be done in 22 languages. Initial experiences are underway in English and Indian.

“It will be very dedicated, and it will not be away from the shelf,” he says.

Returning to the city’s streets, Vineet Sawant increased the number of parcels it provides from ten to 30 per day, and partially assisted of the translation function in the delivery application.

He believes that he will help more people like him.

“This makes us feel that we belong. Not everyone understands the English language. When the application speaks our language, we feel more confident, and we work better.”