Intents

Let's say someone tells you the words "I am having an issue with one of your products." What are they communicating with those words? They are saying they have a problem and are looking for help. That is the intent of their sentence. For Bavard assistants, intents are pre-defined labels that can be used to categorize what a user is saying to a chatbot. The labels (or intents) can be whatever you want them to be: "needs_help," "saying_hello," "asking_for_human," "asking_for_prices," etc. Your Bavard assistant's intents can be specific to your domain. They allow you to automatically identify what your chatbot user wants, or is trying to communicate, which in turn allows you to then give useful responses to address your user's wants. These responses are called actions. You can help your assistant know which actions to take by defining them in a dialogue policy, or by giving examples in the form of training conversations.

To get your assistant to start recognizing intents, go to the Intents tab of the NLP section of the Bavard web app. Once there, define your intents, then provide a few natural language examples for each one. For example, given the intent "ask_for_prices," some natural language examples might be:

  • "What are your prices like?"
  • "How much does x cost?"
  • "Are your services expensive?"

The example "How much does x cost?" is a special case. Here, "x" is likely the name of some product. Bavard assistants can be used to recognize and extract those (we call them tags), so you can react to that information intelligently.