# ShinrAI 1.3 language and category coverage

Business and clinical letters; 200 texts per locale. Strict entity spans. Hebrew beta is included. Model-only evaluation by Innovius, not a service-wide accuracy guarantee.

## Model locales

- German (de): available; 94,4% strict-span F1.
- English (en): available; 97,7% strict-span F1.
- Japanese (ja): available; 98,1% strict-span F1.
- French (fr): available; 96,0% strict-span F1.
- Spanish (es): available; 96,9% strict-span F1.
- Italian (it): available; 95,0% strict-span F1.
- Polish (pl): available; 90,9% strict-span F1.
- Portuguese · Portugal (pt-PT): available; 94,2% strict-span F1.
- Portuguese · Brazil (pt-BR): available; 97,3% strict-span F1.
- Russian (ru): available; 96,4% strict-span F1.
- Ukrainian (uk): available; 86,7% strict-span F1.
- Turkish (tr): available; 90,1% strict-span F1.
- Arabic (ar): available; 87,9% strict-span F1.
- Korean (ko): available; 95,5% strict-span F1.
- Hebrew (he): beta; 59,5% strict-span F1.

15 locales include two Portuguese regional variants. Website translations: en, de, ja, fr, es. Azure text adapter language codes: de en fr es it pl pt ru uk tr ar he ja ko. API acceptance is separate from category coverage and quality.

## Categories

### People and organisations

- People (`PERSON`); example: `Emma Weber`.
- Organisations (`ORG`); example: `Example GmbH`.

### Contact and location

- Cities (`CITY`); example: `Berlin`.
- Street addresses (`STREET`); example: `Musterstraße 12`.
- Email addresses (`EMAIL`); example: `emma@example.com`.
- Phone numbers (`PHONE`); example: `+49 30 000000`.
- Postal codes (`POSTAL_CODE`); example: `10115`.

### Finance

- Bank accounts (`ACCOUNT`); example: `DE89 3704 0044 0532 0130 00`.
- Crypto wallets (`WALLET`); example: `0x…`.
- Payment cards (`CARD`); example: `4242 4242 4242 4242`.

### Personal identifiers

- Dates of birth (`DOB`); example: `14.03.1985`.
- National identifiers (`NATIONAL_ID`); example: `Tax / identity / social-security ID`.
- Vehicle plates (`PLATE`); example: `B AB 1234`.
- Customer and case IDs (`CUSTOMER_ID`); example: `CUSTOMER-0042`.

### Digital and confidential information

- Usernames (`USERNAME`); example: `example\emma`.
- Web addresses (`URL`); example: `https://example.com`.
- Network addresses (`NETADDR`); example: `192.0.2.1`.
- Secrets and access tokens (`SECRET`); example: `API_KEY_EXAMPLE`.
- Project codenames (`CODENAME`); example: `Project Aurora`.


### Subtypes

- `PERSON`: common, uncommon, rare.
- `CITY`: major, medium, small.
- `STREET`: generic, specific, local.
- `ORG`: international, national, regional.

The 19 heads contain 27 subtype classes: four heads have three subtypes, and 15 have one. Enterprise deterministic rules and validation are additional and excluded from model-only scores. Compatibility adapters expose only their documented mappings; use the native API for other findings.

[Benchmark source](https://innovius-shinrai-pii-benchmarks.static.hf.space/index.html)