Skip to the tool

Tool

Mask Korean phone, RRN, and email

Korean phone numbers, 13-digit resident-ID patterns, and emails in paste or JSONL strings are replaced with tokens or a partial mask. Nothing leaves this tab.

Matching runs in this tab. GearDel does not receive the paste. The hit list shows line numbers and counts, not the original numbers.

What is matched

Email is a conservative local@domain pattern. It is not a full RFC 5322 parser.

RRN is 13 digits, optional hyphen, first six a YYMMDD with month 01–12 and day 01–31, and a gender digit 1–8. The official checksum is not required, so some other 13-digit ids that look like a date can match.

Phone covers Korean mobiles (01x), area codes (02, 031–064), 070, 080, and 15xx/16xx/18xx national numbers, with or without hyphens. Digits that sit inside a longer number are skipped.

JSONL mode walks string values inside each object, including nested arrays. Keys are not rewritten. A leading JSON array is rejected.

Token style writes [EMAIL], [RRN], [PHONE]. Partial style keeps an email local prefix, last four phone digits, and stars for the RRN. Examples on this page are fictional.

What this page does not do

  • It is not named-entity recognition. Names, addresses, account numbers, and card numbers are out of scope unless they also match the patterns above.
  • A clean result is not a legal finding under PIPA or GDPR. Review the file yourself before training.
  • It does not call an external PII API. At most 20,000 JSONL lines or about two million characters.

Paste

Input shape
Replacement
Patterns

0 phone · 0 RRN · 0 email · 0 lines touched · 0 JSON objects · 0 blank · 0 errors

Questions

Why did an order id get masked?

If thirteen digits look like YYMMDD plus a gender digit 1–8, they match the RRN pattern even without a checksum. Turn RRN off, or split that field before pasting.

Can I share the example link?

Yes. ?example=1 loads the fictional sample on this page. Do not put real resident numbers in a URL.

Does GearDel store the paste?

No. The catalog still does not host datasets. This tool only rewrites text in the tab.

Related on GearDel

Pick a Korean dataset by task →

Korean LLM training rows →

Alpaca vs ShareGPT vs messages →

Eval sets vs training rows →

All browser tools →