Anonymise a variable

This feature allows you to anonymize responses to text-based questions by automatically identifying sensitive information.

List of Anonymized Data

Person	Address	CVC	Company name
Date of Birth	Postal Code	Bank Account Number	Insurance Number
Blood Type	Email Address	IBAN	Username
Medical Condition	Credit Card Number	Social Security Number	IP Address
Medication	Credit Card Expiration Date	Health Insurance ID Number	Social Media ID
Phone Number	CVV	Health Insurance Number	Driver’s License Number
Fax Number	Credit Card Brand	National Health Insurance Number	License Plate Number
Train Ticket Number	Visa Number	Flight Number	Vehicle Registration Number
Booking Number	Organization	Passport Expiration Date	Passport Number
Identity Document Number	Birth Certificate Number	Tax Identification Number	National Identification Number
Student Number	Transaction Number	Serial Number	Registration Number

How does the response anonymization model work?

This feature is based on a named entity recognition (NER) model specifically designed to detect and anonymize sensitive information found in responses to open-ended questions. Unlike Large Language Models (LLMs), this model performs no generative processing, and no data is transmitted to external services.

The model is hosted exclusively on our secure infrastructure, ensuring full control over data processing. No data is sent outside of the iQ3 environment.

Supported languages

French, English, Spanish, Portuguese, German, Russian, Arabic, Italian, Ukrainian, Norwegian, Swedish, and Danish.

NOTE: Anonymization applies to all observations of a variable, regardless of the language in which they were entered.

To anonymize the responses for a variable, proceed as follows:

1. (Required) Log into your SphinxOnline account from Sphinx IQ3.

   Procedure for logging into the SphinxOnline account from Sphinx IQ3 :

   1. Open the Sphinx IQ3 software program.

   2. Click on Connection in the top right-hand corner of the screen.

   3. Select the Server URL (where your account is hosted).

      The drop-down menu allows you to select the server on which your survey will be published. By default, the SphinxOnline (www.sphinxonline.com) shared server is selected.

      

      If the server URL is not present in the list :

      1. Click on the + button.

      2. Select Other server from the Server drop-down menu.

      3. Insert the URL address of your server.

         

      4. Click on Save to validate.

   4. Click on Next.

   5. SphinxOnline will open in your browser: if you are not yet logged in, enter your login details and click on Log in.

   6. A dialog box will open in your browser. Click Open Sphinx iQ3.

      

      NOTE: Check the “Always allow [server URL] to open links of this type in the associated app” option to avoid displaying this window on subsequent logins.

   7. Return to the Sphinx iQ3 application.

      NOTE: Connection to your online account from iQ remains active for a sliding period of 5 days: each new interaction with the server (publication, import, etc.) resets this period for a duration of 5 days from the action.

   Once logged in, the account name appears in the top right-hand corner of the screen.

   

   By clicking on it, you can:

   • Open the online account in the browser.

   • Change account (if the personal account is linked to one or more collaborative accounts).

   • Log out.

2. In the Texts tab, click on Anonymise.

   

3. Select the text variable to be anonymized.

4. Specify the minimum observation length to consider in number of characters (e.g. 2 characters).

5. (Optional) Type in the text to display instead of sensitive data (e.g. XXXXX). 

6. Select where to insert the new variable:

   • Create the variable after the selected variable;

   • Create the variable in the section (section to determine);

   • Create the variable at the end of the questionnaire.

   • If you already anonymized a variable, you can Fill the variable. This can be useful if a respondent changed their response or if new observations were entered and are not yet anonymized.

     NOTE: By default, anonymization only impacts empty cells of the previously anonymized variable. You can choose to Send already anonymized data (i.e. the whole dataset).

7. Click on Anonymise to finish.