Limitations
- You can add up to 2,000 example sentences per intent.
- The minimum number of example sentences is five. If you attempt to train an intent with fewer than five example sentences, a red indicator will appear once the build completes.
Manage example sentences
Create example sentences
Create example sentences
- Manually
- Automatically (generative AI)
To create an example sentence, go to Advanced > NLU, select the flow, and open the intent on the Intents tab.
Enter the example sentence and press Enter or Tab.
Delete example sentences
Delete example sentences
To delete example sentences, hover over a sentence and click 
.
Recommended number of example sentences
For optimal performance, use five to 100 example sentences. With this range, you can provide the agent with enough data to recognize the intent and avoid overcomplicating the model with unnecessary examples. If you don’t meet the minimum of five example sentences or exceed the recommended maximum of 100, Cognigy shows a warning icon prompting you to adjust the number of sentences.Tips for writing effective example sentences
- Avoid punctuation marks. Keep the sentences simple and without punctuation. Using punctuation marks in example sentences, such as commas and question marks, can reduce recognition accuracy.
- Avoid repetition. Don’t write similar sentences like
I want to order PizzaandI want to order Pasta. Instead, use lexicons to group similar terms and attach them to your flow. This way, you can create a more general sentence likeI want to order food, which will automatically recognizePizzaandPastaas food-related items. - Use natural language instead of placeholders. Avoid placeholder variables like
LEX_SLOT_123and limit uppercase usage, for example, don’t writeFOODin all caps unless it’s an abbreviation. - Ensure uniqueness. Make sure your example sentences are different from each other. If they’re too similar, the agent may struggle to differentiate intents, which could cause incorrect results or frequently trigger confirmation sentences.