The magic of a chatbot happens when a customer asks a question using the customer’s own words, and the chatbot understands those words and answers the question correctly. But how does that magic happen? Determining that two questions mean the same thing, even if completely different words are used, is part of the challenge of building a chatbot.
In this talk, I will describe how we developed a tool to support our chatbot designers’ need for training data in our customers’ own words, and how we designed a method of having our customers label data for us, while improving the customer experience at the same time.
