Policybazaar, IISc Come Together to Develop Automated Speech Recognition Algorithms to Effectively Address Consumer Needs
Policybazaar, one of India’s leading online market places for insurance, has collaborated with the Indian Institute of Science to work together on Automated Speech Recognition algorithms to enhance customer experience. In the first-ever initiative of its kind, both parties have signed a contract to develop algorithms that can accurately interpret human speech. With these algorithms, the company intends to leverage meaningful insights derived from millions of calls between customers and its advisors.
As part of the project, Policybazaar would work together with Dr A.P. Prathosh, Assistant Professor at Department of Electrical Communications Engineering, IISc, to build deep expertise in automated speech recognition (ASR) algorithms, and improve the accuracy of the current ASR algorithms and develop out-of-domain adaptations. Dr Prathosh, who has earlier worked at IIT-Delhi, comes with expertise in Deep Representation Learning and Deep Generative Models. He has also co-founded a start-up that builds learning algorithms for behavioral healthcare.
Santosh Bhat, Head of Data Science at Policybazaar.com, said, “As the accuracy of the ASR engine improves, it would improve our downstream deep learning models for NLU (Natural Language Understanding). This will, in turn, help us get to the core of customer issues and provide them with the best possible solutions. Advancement in deep learning has made it possible to explore the state-of-the-art solutions that go beyond the run-of-the-mill approach and we want to leverage it to understand our customers better and eventually serve them better.”
Policybazaar handles a volume of more than 1.5 lakh interactions via calls between its customers and advisors every day. These conversations include clearing up doubts regarding buying a new insurance product or queries related to existing policies. The idea of this collaboration is to analyze these millions of conversations to gain relevant insights into the customer needs, in order to improve advisor productivity and eventually enhance customer experience. ASR is the technology that enables the recognition and translation of spoken language into text by computers. Through this collaboration, Policybazaar also aims to build state-of-the-art algorithms for speech recognition of Indian languages in a robust manner. This will help address the uncommon quantum of customer needs effectively and in turn, help with better consumer engagement.
