Dictionary and exception list

The dictionary

The dictionary (/admin/masking-dictionary) is the central list of terms and patterns that are redacted or pseudonymised automatically in a project’s documents. Select the project at the top:

For each entry the list shows: the Active switch (inactive entries are ignored when redacting), the pattern, a replacement (pseudonym) where applicable, the action (with its type), the style, the tolerance and the creation date. The search field and — where pseudonymisation is licensed — the Action filter (all / redaction / pseudonymisation) narrow the display; the checkboxes allow multiple selection and “Delete selection (n)”.

Creating entries

“New entry” opens the dialog — several patterns can be created in one go:

Field	Meaning
Action	The Redaction group: “Redact word”, “Redact line”, “Redact sentence”, “Redact paragraph”, “Redact document” · the Pseudonymisation group: “Pseudonymise”. The groups only appear with the corresponding licence.
Style	For redaction: Filled (with a choice of colour), Highlight or Blur; for pseudonymisation: No marking or Highlight.
Type (per pattern row)	Word (a single word), Word pair (“Mona Muster”, for example), Pattern (a regular expression) or Pattern (line) (a regex that redacts the whole line).
Tolerance	0–5 (for word and word pair only): the permitted deviation for matches that tolerate typos and OCR errors.
Pseudonym	Only with the “Pseudonymise” action: the substitute word. If the field is left empty, the system proposes a suitable pseudonym automatically.

Examples:

• Redact every occurrence of “Mustermann”: type Word, pattern Mustermann, tolerance 1 (which also catches OCR errors such as “Musterrnann”).
• Replace “Mona Muster” with “Edgar Siebendar”: action Pseudonymise, type Word pair, pattern Mona Muster, pseudonym Edgar Siebendar.
• Redact IBAN lines: type Pattern (line), pattern \bAT\d{2}[ ]?(\d{4}[ ]?){4}\b.

Ambiguous pseudonyms

If you assign the same pseudonym to different originals, the page warns you (“Pseudonym assigned more than once for different originals — this makes re-identification ambiguous”) — resolve that before exporting, otherwise the reverse translation is no longer unambiguous.

Tools: import, export, adoption

Besides “New entry”, “Find data” leads to the project’s automatic PII detection (matches found can be adopted into the dictionary). The “⋮” (advanced functions) menu gathers:

Function	Purpose
Area rules	Redact fixed areas of the page (a letterhead, for example) project-wide — a page of its own within the project.
Import AD	Adopt personal names from the directory service as dictionary entries.
Import JSON / Export JSON	Back up or apply the dictionary as a JSON file (max. 10 MB; with a duplicate preview and an “Overwrite duplicates” option).
Mapping (CSV) / Mapping (JSON)	The original-to-pseudonym mapping for an external handover (re-identification or auditing outside the system) — only where pseudonymisation is licensed.
Adopt from project	Copy the dictionary and/or exception list of another project — in Merge mode (skipping duplicates) or Replace mode (emptying the target first).
Delete all	Empty the project’s dictionary completely (with a confirmation prompt).

The exception list

The exception list (in the project area: /projects/{id}/masking-whitelist) is the counterpart to the dictionary: words and expressions listed here are not redacted, even when the dictionary or the tolerance search matches them — useful for company names or technical terms that look like personal names.

Entries are stored automatically as a word or a relation (word pair), depending on their form; import and export as JSON is possible. Words marked manually in the document viewer always take precedence and are redacted despite the exception list.

In practice

If your own company name “Muster GmbH” wrongly appears in redaction proposals, enter it in the exception list — detection will skip it project-wide from then on.