Slot-filling is the key capability of Power Virtual Agents that ensures a smooth, natural conversation between the chatbot and end-user. For example, if a user provides a piece of information before the bot has asked for it in a question, the bot is able to recognise this. Hence, the bot understands the users intent, and will not ask the question.
To see this work in action, let’s take a look at an example. In this topic, the bot will be asking the user a series of questions to capture information about a property the user is interested in inspecting. The questions and entity being captured for each are:
Are you looking to rent or buy?
Multiple Choice: Rent, Buy
What state are you interested in?
Custom entity for Australian States: NSW, VIC, QLD
How many bedrooms are you interested in?
You can see above, the user has responded to each question as the bot has asked it. But a more natural conversation might go like this: In response to the first question, “Are you looking to rent or buy?”, the user states upfront the message: I’m looking to buy a new, double storey house with 4 bedrooms in NSW.
Rather than follow the conversation tree step-by-step and ask all the questions, the bot will recognise that an answer has already been provided. It will then proceed to skip the questions, and move ahead in the conversation.