Detection classes

The entity_type in a response is one of the values below, or {"other": "<category>"} for a custom rule. The token label is the class in upper case, so national_id becomes NATIONAL_ID_wbqtrz.

Identity

Class	Found by	Validation
person	Entity model, then verifier	Name shape; junk filtered by the normaliser lexicon
organization	Entity model, then verifier
national_id	Rule	TCKN two-step checksum, first digit non-zero. Also accepts a separated form, and 8 to 12 digits next to a “T.C. Kimlik” or “TCKN” label without a checksum, for OCR-damaged values. Foreigner IDs (99 + 9 digits) under a label.
passport	Rule, label-gated	Shape under a passport caption. Machine-readable zones are recognised by their check digits.
date_of_birth	Rule under a birth-date label or role word, and the role model	Year between 1900 and now. A bare date is never a date of birth.
place_of_birth	Rule under a caption, and the normaliser retyping a location
nationality	Rule under a caption, and the normaliser retyping a location
license_plate	Rule	Turkish plate grammar, province 01 to 81, with a plate word nearby

Contact and location

Class	Found by	Validation
phone	Rules	Numbering-plan check with Türkiye as the default region; twelve further regions with a country prefix or label. Excludes values that are a valid TCKN, a Luhn-valid card or a date.
email	Rule	Address shape with word boundaries
address	Normaliser, composed from a model location plus unit nouns and numbers
location	Entity model, then verifier
coordinates	Rule	Decimal pairs inside the Turkish bounding box, or anywhere under a position label; DMS and hemisphere forms
url	Rule	www. or an http(s) scheme
ip_address	Rule	IPv4 and IPv6 parsed as addresses

Financial and account

Class	Found by	Validation
iban	Rule	ISO 13616 mod-97, length 15 to 34, Turkish and foreign
credit_card	Rule	Luhn, four groups of four. A card inside an IBAN is ignored.
vkn	Rule, label-gated	Ten digits under a tax-number label; the caption-above layout also requires the VKN check digit
bank_account	Rule, label-gated	Account numbers under a label; SWIFT/BIC with an ISO country code
customer_id	Rule, label-gated	6 to 20 digits under a customer or subscriber label; excludes valid TCKNs and cards

Telecom and device

Class	Found by	Validation
imsi	Rule	286 plus 12 digits bare; 14 to 15 digits under an IMSI label
iccid	Rule	89 plus 17 or 18 digits, Luhn
imei	Rule	15 digits, Luhn; IMEISV as 16 digits under a label

Custom categories

Rules published in the management console add categories of your own, per tenant or globally: case numbers, policy numbers, internal identifiers. A rule is a regular expression with a category, a confidence and a lifecycle (draft, active, disabled). Published rules take effect at the next request. A category that is not a built-in class is reported as {"other": "<category>"} and tokenised under its own label.

Languages

Rules carry Turkish and English labels throughout, with German phone formats and French captions for passports, place of birth and nationality. The entity model is trained on Turkish. Identifier rules are Turkish-specific where the identifier is (TCKN, VKN, plates) and international where it is not (IBAN, cards, passports, IP addresses).