European Work Information with Redacted Personal Identifiers

50 sample entries provide a preview of the PII-Masking-2M corpus, focusing on European work and HR information where personally identifiable information has been redacted. The full dataset contains 200,000 records, created by AI4Privacy and released in April 2026. This preview demonstrates the data structure and label distribution without exposing the original sensitive text.

Use Cases

• Train PII detection models using the provided labels for categories like names, addresses, and IDs on synthetic work-related text.
• Benchmark the performance of text anonymization tools against the ground-truth PII labels in the dataset.
• Analyze the distribution of PII label types (e.g., person, organization, date) across European work documents.
• Generate synthetic training data for downstream NLP tasks by using the redacted text as a template.
• Study language distribution patterns within professional and HR contexts across different European regions.

Strengths

• Part of a larger 2-million-record corpus (PII-Masking-2M).
• Provides explicit label and language distribution metadata for analysis.
• Focuses on a specific domain (European work and HR information).

Limitations

• Preview contains only 50 sample entries, insufficient for model training.
• Source text and actual PII values are redacted, limiting direct use for some tasks.
• Geographic coverage is specified as European, but specific countries are unknown.

Provenance

Source

AI4Privacy.

Collection Method

Presumably synthetic or processed from real-world text with PII redacted; exact method is null.

Time Range

null

Freshness

Released or updated in April 2026.

Geography

European coverage.