AI startup Devnagri offers accurate translations in 22 Indian languages | Fusion - WeRIndia

Inspired by the versatility and diversity of the Devanagari script, Himanshu Sharma and Nakul Kundra founded Delhi-based startup Devnagri. Devanagari is the fourth most widely used script in the world,

Devnagri is a human translation platform which uses a combination of machine translation and machine learning with human intervention to provide accurate and contextual translation in 22 Indian languages.

The startup primarily helps businesses translate their software communication tools at scale into Indian languages processing millions of words every day. Even though there are various translation tools available, Devnagri claims to be ten times more efficient and accurate than a traditional translation platform due to human mediation along with AI in the translation process.

The platform uses end-to-end open-source machine learning by Google called TensorFlow. They built their own proprietary translation engine to provide a rough translation.

The rough translation is then processed by human intervention to make up for the absence of context. Feedback is provided to improve the accuracy of the translation with every run.

Devnagri offers three types of products:

Translation-Clients who upload their text into the platform which is then translated using AI and further improved upon by human translators.

Devnagri on the Air (DOTA), a JavaScript plugin that the client can use in any application or browser to enable real-time translation without any necessity coding.

And lastly, Devnagri Transliteration Application Programming Interface which translates names and addresses into Indian languages and avoids proper nouns being to other languages.

Kutumbh is Devnagri’s community of 5000 translators who know a minimum of two Indian languages along with English. They, along with 16 employees, form the backbone of the organization.

Devnagri has 30 clients all over the world and converts text from many international languages like English, Japanese, Chinese etc.

Image Credit: Devnagri Website Screenshot

Image Reference: https://devnagri.com/