You can add utterances manually to any intent by typing them directly into the intent. Utterances added via intent sets and manually all count towards the total number of utterances.
Required Editions
| View supported editions. |
| User Permissions Needed | |
|---|---|
| To build and manage Einstein Bots: |
Customize Application OR Modify Metadata OR Manage Bots |
| To manage intent sets and bot training: |
Customize Application OR Modify Metadata OR Manage Bots OR Manage Bots Training Data |
Before you begin, keep these rules in mind when writing utterances:
- Building an intent model with Einstein Bots is supported in the following languages: Create a Bot for a Specific Language.
- The minimum number of utterances required to use intent recognition varies by model.
Einstein Bots offers two intent models, the original intent model and the cross-lingual intent model (beta). Learn more about the models and the minimum requirements. Before you meet the minimum requirements, both models default to exact matching. After you meet the minimum requirements, you can turn on Einstein to use Natural Language Processing for the intent.
- Utterances over 250 characters are discarded.
- The more balanced your model, the better your model accuracy.
In many cases, the more utterances, the more accurate your model. But keep in mind that too many or too few utterances can change the model output. The minimum number of utterances depends on the model you use, and the right number of utterances is unique to your model.
- Utterances must have variation in word length.
The words must vary in length to capture the many different ways an end user could ask for something. Some end users could say “order status”, while others say, “I would like to check the status of my order”. A great way to create utterances is to ask your colleagues how they would say a phrase.
- Avoid adding utterances that are too similar.
Also avoid adding the same utterance more than one time, as the model discards duplicate utterances.
- Special formatting for utterances isn’t necessary.
Special text formatting, such as emoji, words in all uppercase, and punctuation aren’t included in the training process.
- Each utterance is clear and includes the activity that’s intended.
Avoid utterances that have an ambiguous meaning. An intent-driven utterance follows this template to be descriptive:
- Includes an actionable verb or task. For example: "play," "display," or "switch on/off."
- Includes a specific entity or noun to be acted upon. For example: "Silicon Valley episode," "bank transactions", or "weather."
- Navigate to the Bot Builder, and select the dialog you want to add utterances to the intent.
- Under Utterances, type in your utterance, and click Add.
- Repeat for another utterance.