Improving your domain model
The Learning page helps you to train the domain model by improving the accuracy of matching intents to slots.
The Learning page lists utterances that were not matched to an intent. You can improve your model by assigning an utterance to an intent. You can also delete unrequired utterances.
Assign utterances to intents
- Select an utterance and click Confirm under the Predicted Intent column to add that utterance to the intent.
Alternatively, you can choose one or more utterances and select an intent that the utterance relates to in the Assign to Intent option.
The utterance will be mapped to the predicted intent.
The utterance is moved to the intent and can now be mapped to slots from the intent's Utterances tab.
Delete unrelated utterances
- Choose an utterance from the workspace and click Delete.The utterance is removed from the workspace.
You can filter the list of utterances by status or by intent.
- Click the Filter by Status option and select a status filter:
- All - displays all utterances.
- Incorrect - displays utterances that were mapped to an intent but the match could be incorrect.
- Correct - displays utterances where the intent was identified correctly.
- Unknown - displays utterances that the bot can't correctly map to the intent.
- Click the Filter by Intent drop-down and select an intent filter:
- All - displays all utterances
- None - displays utterances where the bot could not predict an intent.
You can also search the utterances for a specific keyword using the Search field.