Insights

Overview

Insights aid in fine-tuning your redaction approach by recognizing crucial occurrences that may arise during a conversation. This, in turn, ensures that you avoid indiscriminately redacting information, as it considers both the specific queries posed and the overall context of the conversation.

Insights can be used works in conjunction with the out-of-the-box functionality provided by the NLP engine.

Insights Details

The details section sets the base parameters for the insight.

Configuration

Trigger

Triggers are specific occurrences that initiate the activation of an insight. For example, one such trigger could be when an agent requests the customer's social security number.

Response

Responses correspond to the actions resulting from an insight trigger. In the case described earlier, the customer's act of providing their social security number to the agent would be considered a response.

To validate a triggered insight, make sure to set the "validation" field to "true" within the configuration.

Expression Validation

Validation prevents insights with critical expression flaws from executing and resulting in data leakage.

Expressions can be validated by calling the the validation endpoint.

Configure Redaction

Sets the redaction options for both legs within a call.

The setting below only redacts the text from the customer's utterance and not within the audio file.

{
    "response": {
        "expression": "\\b\\d{3,5}\\b",
        "before": 2,
        "after": 2,
        "redaction": {
            "type": "all",
            "audio": false,
            "text": true
        }
    }
}