Models

Names are the part of language that does not generalise, so the models that read them are trained on recorded names rather than on running text. These are the ones this site runs on.

  • Parse

    Try it →

    Splits a full name into labelled parts — given, surname, patronymic, particle, title — across writing systems and naming conventions, and returns alternative parses where the string is genuinely ambiguous.

    POST /parse

  • Transliterate

    Try it →

    Romanises a name from 24 scripts, and restores diacritics on names that lost them. Thai merges two models — a corpus-attested graphone route and a neural model for coverage — and every variant carries a confidence.

    POST /transliterate

  • Classify

    Live as a service; no demo page yet.

    Given a name, predicts what it is: the language and country it belongs to, whether it is a given name or a surname, and the gender it is recorded with.

    POST /classify

  • Name graph

    Try it →

    The spellings, scripts, languages, countries, relatives and register evidence for one name — the data every name page on this site is built from.

    GET /name-graph/{form}

  • Evidence

    Try it →

    Which official registers attest a spelling, which reference works list it, and one register’s count over time. Licensed figures never leave the lake: books and name dictionaries come back as presence, never as a number.

    GET /name/{form}/evidence

Calling them

Every endpoint above is on https://api.mondonomo.ai/api/v1. The OpenAPI description covers the public surface; get in touch for volume, licensing, or a field this site does not publish.