PII Filtering

This activity helps to identify and redact PII from the provided input text to ensure compliance with data protection regulations.

Technical Reference

Configuration

Once the integration is connected from the Manager, the property panel will automatically display the connected integration.
If you see the message “Add new connection”, click here to learn how to set up a new connection.

Below are the properties available after the project has been integrated:

Provider: This parameter indicates the account name associated with integration.

Model Type: Specifies to choose the list of models available for GEN AI integration. 

Input

DelayAfter: It assists the user to add a delay before initiating subsequent activities. The delay duration here is in milliseconds. By default, it is set to “500” milliseconds. When the option is left blank, the delay will not be considered.

DelayBefore: It assists the user in adding a delay before starting the execution of the activities. The delay duration here is in milliseconds. By default, it is set to “500” milliseconds. When the option is left blank, the delay will not be considered.

Input Text: *Specifies the input text in which PII needs to be identified and redacted and it accepts values in String datatype.

PII/PHI Category: *Specifies the PII/PHI categories to be identified and redacted from the input text. Multiple categories can be selected from the drop-down.

Timeout: Specifies the maximum time allowed for the activity to execute. If the connection is not established within this period, an exception will be thrown. By default, it is set to “30000” milliseconds. 

Test: Selecting this option opens the Co-Pilot assistant, allowing you test the provided input and view the output simultaneously.

MISC

DisplayName: Displays the name of the activity. The activity name can be customized, which aids in troubleshooting.

SkipOnError: Specify the “Boolean” value as “True” or “False.”
True: Continue executing the workflow regardless of any errors thrown.
False: Halt the workflow if it encounters any errors.
None: If the option is left blank, the activity will, by default, behave as if “False” were chosen.

Version: It indicates the version of the feature being used. 

Option

Minimum Confidence Score: Specifies the minimum confidence score required for the activity to identify and redact PII data from the input text. The score should be a value between 0 and 1 (e.g., 0.85).

Text Language: Specifies the language of the input text. Select the appropriate language from the drop-down to enable accurate detection and redaction of PII data.

OUTPUT

Redacted Text: Returns the redacted version of the input text based on the selected PII category and the output is returned in a “String” datatype.

Result: It provides the ability to view the execution status of the activity. It returns values in “Boolean.”
True: Indicates that the activity has been executed successfully without any errors.
False: Indicates that the activity has been unsuccessful due to an unexpected error being thrown.

* Represents mandatory fields to execute the workflow.

Understanding Confidence Score

The confidence score sets the minimum certainty the activity needs to identify and redact PII from your text. It helps avoid unnecessary redactions by filtering out low-confidence results.

You can adjust how sensitive the redaction should be:

a. High score (e.g., 0.9 to 1): Redacts only when the AI is very sure and ensures accurate redactions.
b. Low score (e.g., 0.6 to 0.9): Redacts more, but may include some incorrect redactions.