Smart highlights automatically detects highlights in a contact center conversation without requiring any additional configuration. A conversation's highlights are the words or phrases that were important for determining the end user's intent. This feature analyzes each conversation turn and categorizes the user's intention, then labels any highlights present in a conversation.
How it works
Conversational Insights uses Dialogflow intent technology to detect conversation highlights at the sentence level. When Conversational Insights detects a highlight, it creates a call annotation and includes that in the analysis with other annotations.
Supported smart highlights
| Intent | Definition | Example highlight |
|---|---|---|
AUTHENTICATION_INFO |
Asking for information that can be used to authenticate a customer. | Can you please help me with your account number? |
CHECK_ISSUE_RESOLVED |
Checking if the issue is resolved and there are no further questions or problems. | Have I addressed your concern for today? |
CONFIRM_ISSUE_RESOLVED |
Confirming the issue was resolved. | All my questions were answered. |
View smart highlights
You can view smart highlights labels in the Conversational Insights console along with other labels, such as detected entities. When you export conversations to BigQuery, Conversational Insights automatically includes smart highlights labels in your exporter. This lets you perform a custom inspection of the data.
Smart highlights feedback
Smart highlights use natural language processing to identify intents within your conversation. This feature can incorrectly identify phrases or not identify an accurate phrase. If you find a scenario in your data where a smart highlight is not working as described, you can provide feedback with one of the following methods:
- Use the Send Feedback option from the Help menu in the Conversational Insights console.
- Reach out directly to the Conversational Insights feedback team.
In your feedback, provide the name of the smart highlight, the phrase, and whether the phrase was missed or identified incorrectly.