MODEL 02 · CROSS-SCRIPT TRANSLITERATION

Transliterator

      
      Powered by MondoPhone · Thai uses AyutthayaAlpha today
    

Convert names between writing systems — corpus-weighted, not naïve character mapping. The model knows that Yousuf, Yusuf, Yousef, Youssef and Yusef are all valid Latinizations of يوسف — and which one your Saudi customer is most likely to use. For Thai-Latin specifically, the production model is AyutthayaAlpha (CER 0.0047).

Model card REST API docs

Try it — name → 8 writing systems

Enter a name in any script. The model returns its most-common form in eight world writing systems, ranked by corpus frequency where multiple variants exist.

EXAMPLES:

Enter a name above and press Transliterate.

Model card

Approach: Corpus-weighted lookup + PolyIPA fallback
Scripts covered: 150+ (Unicode-script-tagged)
Corpus: Nomograph DB — 170M+ names with multi-script occurrences
Fallback path: Grapheme → IPA (PolyIPA) → target-script G2P
Romanization standards: RTGS (Thai), ALA-LC, BGN/PCGN, ISO 233 (Arabic), Hepburn (JP), Pinyin (CN), Revised Romanization (KR)
Latency: ~25ms typical · single batched call returns all targets

API

curl -X POST https://api.mondonomo.ai/v1/transliterate \
  -H "Authorization: Bearer $TOKEN" \
  -d '{
    "name": "Eugen",
    "source_lang": "de",
    "targets": ["latn", "cyrl", "arab", "thai", "jpan", "hans", "deva", "hebr"]
  }'

# Response
{
  "results": [
    {"script": "latn", "primary": "Eugen", "variants": ["Eugene", "Eugène"]},
    {"script": "cyrl", "primary": "Юджин", "variants": ["Євген"]},
    ...
  ]
}

Why not just use a Romanization standard? Standards like RTGS or BGN/PCGN are prescriptive: they say what the spelling should be. Real names from real corpora are descriptive: people spell themselves how they want, and the chosen form is a strong signal of identity. The Transliterator returns the descriptive distribution, with the prescriptive form available as a tagged variant.

MODEL 01

PolyIPA

The phonetic backbone when no direct cross-script corpus exists.

g2p · p2g→

MODEL 06

Soundalike Search

Find names that sound alike across the scripts you've just transliterated to.

echoes→

MODEL 04

Name Parser

Parse first, then transliterate each part — particle handling varies by culture.

structured→

Transliterator

Try it — name → 8 writing systems

Model card

API

Models that combine well.

PolyIPA

Soundalike Search

Name Parser