3 in 5 Americans would try a new brand or company for a better service experience, and yet 78% of American consumers have bailed on a transaction because of a poor service experience. This means that good customer service should be a central part of any business's money-making plan, yet the majority of them succumb to the pitfalls of bad customer service and lose money. In fact, many companies lose more than 600 wage earning hours to inefficient customer service practices every year. In order to solve this problem, we attempted to see how to improve the efficiency of customer service while keeping the quality and level of customer appeasement high. Generally speaking, this is a paradoxical problem. People hate to be put on the phone with bots, but human customer service representatives are extremely inefficient. After some thinking, we came to the solution of making bots that are as human as possible.
What it does
ILOY uses multiple different datasets in order to recreate the users voice and tenancies of speaking during a phone call. It uses recordings of the users voice compiled with Festival Speech Synthesis to create a voice that sounds like the user. It then trains a recursive neural network with the users regular daily speech which will allow ILOY to interact with other people in the same way that the original person would in real life. The final step is to plug this into a telephone service which allows ILOY to act and sound like the original user. With all of these things, ILOY will be able to create a convincing simulation of the user in speech and text.