Setting up Natural Language Understanding (NLU)

Note: To add intents to the EngageOne™ Communicate tenant (company account), click Bot Settings, then select NLU under Language.

You can extend the range of utterances for an intent by training the bot to use natural language understanding (NLU). This means that the bot can find a match even when the user uses a phrase that is not an exact match for one of the standard utterances.

Adding entities

Entities are separate from intentions. Any intent can use any entity.
  1. In Bot Settings, select NLU (under Language), then go to the Entities tab.
  2. To add an entity, click Create Entity and enter the details:
    Entity name The name or category of the entity, such as restaurant, vehicle, location
    New entity value At least one value for the category.For a restaurant entity, examples might be restaurant, diner, takeaway.
    Note: Entities and their values should not overlap. For example, if you have a restaurant entity and a food entity then do not use pizza as an example for both. It is either a restaurant type or a food type but not both.

    You can add examples for multiple languages.
    New synonyms Values considered to have the same meaning as the entity value. These are optional and are simply a way of organizing the examples. You can enter synonyms in the New entity value field if you prefer.

    For a takeaway example, synonyms might be takeout and drive thru.

    For a food entity where the example is soup then synonyms might be vegetable soup, tomato soup.

Adding intentions

To add an intention for natural language understanding:
  1. In Bot Settings, select NLU (under Language), then go to the Intents tab.
  2. To add an intent, click Create Intent and enter the details - Intent name and Add new utterance. For each intention, you can add a range of utterances as training data. You do not need to add all the messages that users are likely to add, only a sample. For background information, see:

Adding intentions in multiple languages

When adding non-English utterances, you must select the language that matches the language of the bot. The bot supports one language only. If you do not select a language, it defaults to English (en).