About utterances and intentions

Utterances are the responses that users enter into a conversation with a bot. The utterance is the way in which users express themselves. Intentions are what the user wants to communicate or wants to happen.

For example, when a user responds to a bot with:
  • "time" (utterance), they mean "what is the current time?" (intention)
  • "balance" (utterance), they mean "what is my balance?" (intention)
  • "sure" (utterance), they mean "yes" (intention)

Use the Intentions page in Converse to map between what a user says (utterances) to what they may want to communicate or occur (intentions).

Example

Sample intention:

Intention name Utterances
device android
apple
computer
ios
iphone
laptop

When a bot receives a response that matches one of the utterances in the intention, for example laptop, it runs the interaction that is configured with this intention.

For an example of a bot that uses an intention to handle unwanted replies, see Example: handling unwanted and unexpected answers.

How utterances are matched to intentions

A bot looks for a match for utterances in the intentions that are linked to any of its interactions (on the interaction's Details tab). For example if there is a birthday intention with the utterance your age but none of the bot's interactions are linked to the birthday intention then the bot will not understand when the user responds with your age.

When the user responds to a bot with a message containing words that map to different intentions then the bot will always use the first intention it finds. For example:
  • There are two intentions covering utterances for "phone" and "complaint"
  • The bot uses the phone intention when a user sends the message: tell me your phone no - I want to complain
  • The bot uses the complaint intention when a user sends the message: I want to complain - tell me your phone no

Defining intentions and utterances

To set up an intention, you can manually add as many utterances as possible to the intention in order to capture all the ways of expressing something. See Adding intentions.

You can also extend the intention by training the bot to handle a wider range of utterances. See Setting up natural language understanding (NLU).