Intents are components used to detect the intention from user input.
The detection is based on the expression examples that can be added in the detail view.
The detection will match all the expression examples and other similar expressions.
There are two levels of detection. The first one is rule based, combining grammar rules, synonym lists and entity matching. The second level is the NLP level that uses deep learning to understand the general meaning of the example expressions and detect other expressions with similar meaning. It is recommended to have at the very least 10 examples before activating the NLP detection for an intent.
It is recommended that for each intent based ramification you add a fallback intent that will be selected if no other intent was matched. To make an intent the fallback intent check the This is a Default Intent checkbox in the detail view.
There are two more advanced settings in the intent detail view:
- Allow one entity regex allows matching expression examples that only contain one entity. This behavior is disabled by default to avoid misconfigurations that will matching everything.
- Superintent allows grouping multiple intents in a category. This way if the NLP engine can't detect the exact intent but detects that is one of the intents in the category it will match the superintent.