How to get AI to work in your 22 languages

Vineet Sawant has spent the last two years navigating Mumbai streets on a scooter as a delivery driver.

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

But when he started the language barriers were an additional problem.

His first language is Marathi, and G -n SOant speaks “very little” English. “I can understand, but it’s very difficult to read,” he explains.

This created problems for his new job.

He said, “It was difficult at first. Everything was in English and I can understand some of it, but I was more comfortable with Marathi. Before I asked other guys to help me understand what to do.”

His Zepto employer promises “the fastest online supply of food in India”. So drivers fighting the delivery instructions was not perfect.

To smooth this process a year ago, Zepto partnered with Reverie Language Technologies to introduce AI translation service for its drivers.

Since then, his delivery drivers can choose between six languages in the Zepto app.

“I don’t have to guess anymore,” says G -n Soant.

“I would have spent more time reading more and sometimes even made mistakes. Now, if the client wrote” Ring Bell “, I get this instruction in Maratti. So, I don’t have to ask or check again. Everything is clear.”

The difficulties of d -n Soant are common.

“India has 22 official languages and hundreds of dialects,” says Professor Pushpak Bhathacharya of IIT Mumbai, one of the leading experts in India in the use of AI in Indian languages.

“Without technology that understands and speaks these languages, millions are excluded from the digital revolution – especially in education, management, health and banking,” he said.

The deployment of new generative AI systems, such as Chatgpt, has made the task more successful.

Huge amounts of data such as web pages, books or video transcripts are used to train AI.

In wide spoken languages such as Hindi and English, which is relatively easy to obtain, but for others it is difficult.

“The main challenge for creating Indian models is the availability of data. I am talking about refined data. Rough quality data are available. But these data are not very high quality, there is a need for filtration,” says Professor Bhathacharya.

“The problem in India is for many Indian languages, especially for tribal and regional dialects, these data simply do not exist or are not digitized.”

Reverie Language Technologies already uses its translation technology, managed by AI for a number of Indian companies.

Co-founder Vivekananda Pani says that although translation technology facilitates communication, there is “potential for less common dialects to be pushed aside.”

“The challenge will be to ensure that the incredible advantages of linguistic progress governed by AI do not accidentally shrink the rich variety of human language.”

In order to deal with the problem that Professor Bhathachar has Contributed to BhashiniA government project to develop these high quality data sets needed to train AI.

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

It started in 2022, it is a huge endeavor, but it has already made great progress.

Currently, Bhashini hosts 350 AI -based language models that have processed more than a billion tasks.

More than 50 government services work with Bhashini as well as 25 government governments.

For example, Bhashini Tech is used in multilingual public service chatbots and to translate government schemes into local languages.

“Bhashini guarantees the language and cultural performance of India by building India -specific AI models rather than relying on global platforms,” says Amitab Nag, CEO of Digital India, Bhashini Division.

He hopes that within the next two or three years, rural users will have access to voice access to government services, financial instruments and information systems in their native languages.

These data sets aimed at India will hope that one day they will give people developing models based on AI, tools to make it much easier to adapt to the entire population.

Currently, designing any AI program for dealing with complex processes such as healthcare can be extremely challenging.

Kshitij Jadhav, Assistant Professor at Koita Digital Health Center at IIT Mumbai, works on an AI program that would help people give up smoking.

He explains that people at different stages of the process need different tips and usually need a well -trained person to make this assessment.

But there are a limited number of practitioners who can help, especially those who can work in several languages, so Professor Jadhav hopes his AI model can overcome the precipice.

AI “will first identify the type of conversation that a person needs and will ask questions accordingly, will show empathy, emotions,” says Professor Jadhav.

And all this, hopefully, will eventually be done in 22 languages. Initial experiments are underway in English and Hindi.

“He will be very personalized, it won’t be just something from the shelf,” he says.

Back to the streets of the city, Vineet Sawant has increased the number of parcels, which it supplies from ten to about 30 per day, aided in part by the translation feature in the delivery app.

He thinks this will help more people like him.

“It makes us feel like belonging. Not every understanding of English. When the app speaks in our language, we feel more confident and work better.”