Business Cases

Concrete examples of how organisations apply Mondonomo technology and data to real-world name processing challenges.

Identity Resolution

Problem: Matching customer records across systems where the same person's name appears in different scripts, transliterations, or abbreviations.

Outcome: Resolves the same name written in different scripts to one identity, using 300+ transliteration routes across 22 writing systems.

Graphone romanization/deromanization routes; mean out-of-vocabulary exact-match@1 ≈ 0.78, 0.90–0.98 for Cyrillic, Greek, Georgian and Armenian.

PNEUMA-DDMondoPhonWFTS
Try with "Mohammed Al-Rashid" →

Soundalike Name Matching

Problem: Finding names that sound similar across languages for fraud detection, KYC, and sanctions screening.

Outcome: Ranks candidates by phonetic (IPA) distance rather than string edit-distance, so it catches cross-script soundalikes edit-distance misses — Catherine ↔ كاثرين ↔ Caterina.

Cross-lingual homophone index of 4.5M entries; grapheme→IPA g2p + consonant-skeleton blocking, ranked by articulatory-feature distance.

MondoPhon
Try with "Müller" →

Gender Inference

Problem: Inferring gender from a name token for personalisation, form pre-fill, and analytics, without asking the user.

Outcome: Predicts gender directly from the name with a graceful "unknown" fallback on genuinely ambiguous tokens.

PNEUMA v10 gender head ≈ 0.94 accuracy on the held-out multilingual evaluation set.

PNEUMAPNEUMA-DD
Try with "Andrea" →

Data Quality Assessment

Problem: Detecting malformed, reversed, or culturally misclassified name fields in CRM and marketing databases.

Outcome: Flags reversed given/surname order, mislabeled organisation and place strings, and encoding corruption via per-token role and script classification.

parse_dict token classifier: 88.8% per-token / 79.3% whole-name accuracy on 136K held-out gold names (2026-07 baseline).

PNEUMAPNEUMA-DD
Try with "García López Ana María" →

Responsible Data Handling

Problem: Classifying whether a data field contains a person's name before applying privacy or retention rules.

Outcome: Distinguishes person names from organisations and places at the token level, so PII rules apply without a full-text NLP pipeline.

Entity-type classification (person / organization / place / other) from the PNEUMA and parse_dict heads.

PNEUMAPNEUMA-DD
Try with "John" →