Yubi Group, a global leader in financial technology, has partnered with AI4Bharat, an AI research initiative from IIT Madras, to develop an Automatic Speech Recognition (ASR) engine specifically designed for handling financial domain conversations in local languages.
This cutting-edge initiative aims to bridge the digital divide by making financial services more accessible to millions of Indians who converse in multiple languages, thus driving true financial inclusion.
The collaboration will focus on developing and training ASR models that can seamlessly recognize and transcribe conversations in multiple languages, including instances where speakers frequently switch between languages, a phenomenon known as “code mixing.”
Yubi’s ASR engine will accurately process such code-mixed statements, ensuring financial institutions can better serve these communities by responding with the right solutions. Initially supporting over 22 Indian languages, the engine will be open-sourced for widespread adoption across industries.
According to the company, ASR engines from tech giants like Google and AWS, fall short in recognising Indian languages and handling code-mixed conversations.
AI4Bharat and Yubi aim to bridge this gap by training the ASR models on a vast dataset, including call center interactions, which represent natural, spontaneous speech in various local languages. The initiative will help Indian businesses optimize customer service and communication through enhanced voice recognition systems and accelerate conversational commerce.
“Financial inclusion begins with conversation. By developing this ASR engine, we plan to enable individuals from every corner of India to engage more easily with financial products and services in their preferred languages. This ASR engine will create a massive impact, allowing businesses to communicate with customers in their native languages, which will drastically improve access to services for those who are currently underserved,” Mathangi Sri Ramachandran, Chief Data Officer, Yubi Group, said.
The ASR models will be integrated into both human-to-human conversations, such as call centre dialogues, and human-machine interactions, such as conversational bots. This will significantly enhance user experiences, enabling real-time voice processing and seamless communication in regional languages.
AI4Bharat’s expertise in natural language processing (NLP) will ensure that the ASR models can handle the complexities of code mixing and semantics between Indian languages. This capability will revolutionize voice-based interactions across industries.