Anonymising documents

Qualitative data often contains personal information: names, employers, locations, dates. Skimle can anonymise or pseudonymise this information for you before your documents are analysed. Anonymisation is configured per data source, so you can sanitise sensitive interview transcripts while leaving other material untouched.

Turning anonymisation on

Each data source has an Anonymise documents switch on its confirmation card, with three modes:

• Off: documents are not anonymised.
• Manual: Skimle detects identifiable information and you review and correct it document by document before analysis.
• Automatic: Skimle anonymises every document automatically using the settings you choose.

When you pick manual or automatic, a settings panel lets you decide exactly how each kind of identifier is treated.

Identifier categories

Skimle detects and groups identifiable information into categories:

• Names
• Titles & roles
• Locations
• Organisations
• Dates & times
• Numbers & codes
• Other

For each category you can choose an action:

• Keep leaves the original text.
• Pseudonymise (coded) replaces it with a consistent code (e.g. PERSON_1).
• Pseudonymise (natural) replaces it with a realistic but fictional stand-in (e.g. a different name).
• Replace swaps in a generic label.
• Redact removes the text entirely.

Cross-file consistency keeps pseudonyms stable across every document in the data source, so the same person is given the same replacement wherever they appear.

Presets

Rather than setting every category by hand, you can apply a preset level:

• Light pseudonymisation: direct identifiers replaced with pseudonyms, with a re-identification key maintained.
• Strong pseudonymisation: direct and indirect identifiers pseudonymised or generalised, with a key maintained.
• Strong anonymisation: all identifiers transformed or suppressed and the re-identification key destroyed. Intended for the strictest compliance needs, with human verification.

You can start from a preset and then adjust individual categories; the settings then show as Customised.

Reviewing anonymised documents

In manual mode, confirmed documents wait in a review step before analysis. Open the data source's Review step to work through each document, checking the detected identifiers and correcting anything the AI missed or over-flagged. Only once you approve the documents do they continue to analysis.

In automatic mode, anonymisation is applied without a review stop, so documents flow straight through to analysis.

When anonymisation runs

Anonymisation happens after you confirm the files in a data source and before they are analysed. If you also have automatic analysis turned on, analysis begins as soon as anonymisation (and any transcription or review) is complete.