Using AI for Hinglish and Benglish

In a country with high levels of illiteracy and multiple languages, the written words are not as effective as spoken ones.

Recognising this trend in India, global and domestic companies are moving towards applying speech technologies for information sharing, messaging application and control systems. The hundreds of millions who are coming online, using mostly smart phones with unrestricted data, find it easier to use their voice and conduct themselves in a familiar language.

The first wave of change occurred when Indian scripts were enabled for websites, phones and apps. This enhanced the user experience of people who didn't or couldn't type in English. A vast number of people were included in the digital mainstream with the widespread use of regional scripts.

The second wave of communication is being enabled by speech software, which is able to understand and apply Indian languages for myriad uses. This helps people who are not used to typing on devices.

MiHup Communications is leading the way on this front from Kolkata, which is emerging as an innovation hub. MiHup is an artificial intelligencebased voice technology platform which "can deliver human-like understanding of naturally spoken queries for large, complex content domains."

A simple example of this is a smart remote. Using MiHup's software, users can give commands to their smart TV remote in their preferred language. The automatic speech recognition software works offline as well as well as with an online connection.

In another useful application, MiHup was able to help citizen engagement body MyGov in receiving voice based messages from several hundred thousand callers who shared their suggestions on a toll free number. The citizens would call up on a certain number and leave a voice message. The MiHup software would convert that voice message into text format and offer it to the organization. The software could analyse the intent of the call and also the topic so that MyGov could take appropriate follow up action. This went beyond a simple transcription to a speech analytics offering. The pitch, tone and volume of the speaker was analysed. If there was a question about issues like lack of utility and job, it would be factored.

This feature has tremendous applications for consumer companies that receive feedback through call centres. For example, about 1.2 million calls are received from consumers per day for a typical large bank. Typically, banks and companies are able to analyse only 1-2 per cent of the calls using manual processes. The MiHup software is able to help analyse every response received in a call centre.

MiHup is working with two leading automobile companies in India to embed voice control software in vehicles. The core voice interface is being customised by the auto companies to help passengers control windows, music and air-conditioning using speech commands. This works in offline mode without connectivity as well.

So far, the company has developed solutions in Hindi, English and Bengali. It has also developed the capability of understanding mixed hybrid languages like Hinglish and Benglish. In the next few months the options in its core offering will be increased to 10 languages. These would be applied in sectors including financial services, health care, transportation, manufacturing, entertainment and smart homes.

Speech-based commands can liberate all consumers and professionals from the tyranny of typing on screens. In many ways, voice would be the main social equaliser to help bridge the digital divide. “The next 700 million digital users in India will use voice and speech for interacting with machines,” says MiHup co-founder and CEO Tapan Barman. "It will democratise the digital experience.

Speech-based commands can liberate all consumers and professionals from the tyranny of typing on screens. In many ways, voice would be the main social equaliser