Persian PII-Masking Combined Corpus: 690K Cleaned Rows

A cleaned corpus of 623,890 persona-clean and 224,956 initial-clean rows for Persian PII-masking token classification. The dataset, created by Reza2kn, combines audited source corpora and was last updated on 2026-05-31. It includes fields such as source_text, masked_text, and privacy_mask labels.

Use Cases

• Train token classification models for PII detection based on the described privacy_mask labels.
• Benchmark PII-masking algorithms on Persian text using the provided source_text and masked_text pairs.
• Develop privacy-enhancing NLP tools for Persian content based on the audited and cleaned corpus.

Strengths

• Combines two audited sources totaling over 690,000 cleaned rows.
• Specific row counts are provided for persona-clean (623,890) and initial-clean (224,956) subsets.
• The dataset is the result of source-specific audits, suggesting a quality control process.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown for the combined split, which may limit suitability assessment.
• Last updated 2026-05-31 04:41:34; freshness should be verified.

Provenance

Source

huggingface

Collection Method

Combination and cleaning of the audited persona-clean and initial-clean corpora.

Freshness

Last updated 2026-05-31 04:41:34.

Geography

Iran/Persian-speaking regions