A Robust Methodology for Building an Artificial Intelligent (AI) Virtual Assistant for Payment Processing
Albert P. Sam, Brijesh Singh, Ananda Swarup Das · 2019
The evolution of business interactions with customers has defined a new paradigm of human-machine interaction through AI enabled conversations. In such scenarios, the machine interacts with the human (also referred to as customers) by applying NLP, NLU and NLG techniques that are used to process, understand and generate. dynamic and rapidly changing course of conversations to keep it meaningful and within context. With the changing landscape, today's NLU and NLP components are supported by the latest deep learning algorithms, making them highly palatable of understanding customers intent, forming a cogent response, or taking an appropriate action than was possible in the recent past. But such sophisticated techniques require good quality of sample examples to learn. In this work, we discuss a step-by-step approach to enhance and enable the NLU behind the AI virtual assistant to robustly interpret customers utterances. As a system for interaction, we developed an AI virtual assistant for a payment process to help customer process the due amount in invoice, which is an integral part of several industries, like Banking, Utilities, Telcos, Retailers, etc. We tested our model by processing sample interactions of which a significant percentage of interactions (approximately 75%) went through finishing the intended task of completing the payment.