There is quite a number of myths and legends about virtual assistants. When reading the news and articles, it seems artificial intelligence can learn about everything and then reproduce anything it has learned. This is somewhat true. If it is not trained with business context and rules, that is not helpful. The same applied to artificial intelligence applied to virtual assistants or chatbots. The assistant is certainly using natural language processing to identify intents, i.e. what the customer of the bot is intending to say. It also uses NLP to identify important information in a user’s sentence, for example a country or a date, what is technically called a slot.
We have created a simple banking on-boarding assistant. The example demonstrates how a bot using the power of AWS Lex and domain knowledge can boost the on-boarding process. It is not using the standard questions and answers model, but a dialogue manager that is capable of keeping the context and act accordingly.
Watch the video and let me know what you think. The video only shows one of the possible dialog paths implemented, but gives an overview of the features.
Note: There is a typo in one of the input, that is intentional to show how Lex can correct automatically and interpret the question asked by the user correctly.