Artificial Intelligence Startup Of Two Delhi Guys Is Offering Accurate Translation In 22 Indian Languages

Artificial intelligence startups are nowadays making a huge splash in the market. Lots of them are doing something very unique and providing better solutions than other already existing platforms.      

‘Devnagari' a Delhi-based AI startup is also doing something very unique and out of the box which is also involving a lot of human intelligence.

As you know that Devnagri script is the fourth highly adopted writing script in the world and 120 plus languages are written in this script.

Getting inspiration from this fact and realizing the troubles while using other translating platforms, Himanshu Sharma and Nakul Kundra founded 'Devnagri', which is an AI and machine learning-powered startup.

This platform works on machine learning and neural machine translation and the support of a community for contextualizing the content. This platform provides translation services in 22 Indian languages.

Now you will be thinking that many efficient translation platforms like- Google translate are already available in the market so what was the need to launch this new one? Basically, Google translate and others are not much efficient, and sometimes when you give input through voice, it is unable to identify, and then you have to type it.

That is the main issue but here it is claimed that Devnagri is at least 10 times more efficient and accurate than any other translation platforms available because of human interposition in the translation process of Devanagari.

Devnagri has been selected in the special category at RAISE 2020, which is the biggest AI solution challenge of Indian startups in the category of natural language processing (NLP).

This platform has been specifically designed for assisting the business in translating their software communication tools into Indian languages from any other language in the world.

Working Method Of Devnagri

In this platform, all the data is already fed and it gives a machine translation of the uploaded content. For building its own proprietary translation engine, Devnagri has used TensorFlow, which is an end-to-end open source machine learning by Google.

Devnagri offers three types of products on its site- translation, real-time translation, and transliteration.

In Devnagri translation, clients can upload their text in any format (.pdf, .txt or any other format) and AI and human translators provide an accurate translation. 

The second one, DOTA (Devnagri On The Air) is it a JavaScript plugin that users can add to the browser or any app and can get the real-time translation of any webpage of software.

The third one, transliteration API (application programming interface) assists in rewriting product list names and addresses in Indian languages. For an instance- for the purpose of address, ' Bansal's Palace' should be transliterated to बंसल्स पैलेस not बंसल का महल.

Other Artificial Intelligence Based Sub-platforms by Devnagri

Kutumbh is another sub-platform by Devnagri and it has a community of nearly 5,000 translators and each translator knows at least two Indian languages along with English.

Projects And Work

Devnagri is basically a bootstrapped platform. When it takes a new project, the system breaks it into smaller parts and distributes it to the translators of a particular language. Hence, per day 1 million plus words are translated on the Devnagri platform with very high accuracy. The translation charges can vary from 50 paise to 2 rupees per word.

The team is now working on a system that will be fully voice-based and that can be used as a data collector for several Indian language dialects. This system can be used to train many voice assistants like Alexa- by Amazon.