How to get AI to work in your 22 languages

Vineet Sawant has spent the last two years to navigate the roads in Mumbai on a scooter as a delivery driver.

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

But when he started language obstacles, there was an additional problem.

His first language is Marathi and Mr. Sawant speaks “very little” English. “I can understand but it is very difficult to read it,” they explain.

This caused problems in his new job.

He said: “First, it was difficult. Everything was in English, and I can understand some of it, but I am more comfortable in Marathi. I used to ask other delivery people to help me find out what to do.”

His employer, Zepto, promises “India’s fastest online grocery delivery”. Therefore, it was not ideal to have drivers struggling with distribution instructions.

To smooth this process a year ago, Zepto participated with Revere Language Technologies to start the AI translation service for its drivers.

Since then its delivery driver is capable of selecting between six languages on the Zapto app.

“I don’t have to guess anymore,” says Mr. Sawant.

“Before that I will take more time to read and sometimes there are some mistakes. Now if the customer writes ‘Ring Bell’, I get that instruction in Marathi. Therefore, I don’t have to ask or check again. All this is clear.”

Mr. Sawant’s difficulties are common.

“There are 22 official languages and hundreds of dialects in India,”, one of India’s leading experts in the use of AI in Indian languages, says Professor Pushpak Bhattacharya, Professor of IIT Mumbai.

“Without technology, which understands and speaks these languages, millions are excluded from the digital revolution – especially in education, governance, healthcare and banking,” he explains.

Like Chatgpt, the rollout of the new generative AI system has made the work more important.

Web pages, books or vast quantities like video tapes are used to train AI.

It is relatively easy to get in widely spoken languages like Hindi and English, but it is more difficult for others.

Professor Bhattacharya says, “The main challenge to create Indian language models. I am talking about sophisticated data. Gargry quality data is available. But this data is not very high quality, it needs to be filtered.”

“The issue in India is for many Indian languages, especially tribal and regional dialects, this data is not only present or digitized.”

Revere Language Technologies are now deploying their AI-Interested Translation Technology for a series of Indian companies.

While co-founder Vivekananda Pani says that while translation technology will make communication easy, “less common dialects are likely to push aside”.

“The challenge must ensure that the amazing benefits of AI-operated language progress do not accidentally reduce the rich diversity of human language.”

Professor Bhattacharya is to help deal with the problem Contributed to BhashiniA government project to develop those high quality datasets required to train AI.

Along with dataset, Bhashini has created AI language models and translation services in 22 languages.

Started in 2022, it is a very large venture, but has already made a lot of progress.

Currently, 350 AI-based language models host which has processed more than one billion tasks.

More than 50 government departments along with Bhashini work with 25 state governments.

For example, Bhashini Tech is used in multi-speaking chatbots for public services and to translate government schemes into local languages.

“Bhashini Division CEO Amitabh Nag says,” Bhashini ensured India’s linguistic and cultural representation by creating an India-specific AI model instead of relying on global platforms.

He hopes that within two or three years rural users will have a voice-capable access to government services, financial equipment and information systems in their native languages.

These India-centric datasets hop that one day people will get tools developing AI-based models so that they can be very easy to adapt them to the entire population.

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

IIT, an associate professor at the Koita Center for Digital Health in Mumbai, Kshitij Jadhav is working on an AI program that will help people quit smoking.

He explains that people in different stages of the process require different advice and they usually require a well-trained human to do that evaluation.

But there are limited numbers of doctors who can help, especially those who can work in many languages, so Professor Jadhav is expecting his AI model to bridge the interval.

Professor Jadhav says, “AI” will identify what kind of conversation the first person will identify and frame questions accordingly, show sympathy, feelings. ,

And all this, hope, will eventually be done in 22 languages. Initial experiments are going on in English and Hindi.

“It will be very customized, it will nothing from the shelf,” they say.

Back to the city streets, Vineet Sawant has increased the number of parcels translating from ten to 30 days in part by translation ceremony in the delivery app.

He thinks that it will help more people like him.

“It makes us feel that we are related. Not everyone understands English. When the app speaks our language, we feel more confident, and we do better.”