# Namsor v3

> Namsor turns any proper name into a high-dimensional cultural vector. Namsor v3 embeddings power fraud detection, compliance screening, churn prediction, and audience expansion at global scale.

Namsor is the global standard in morphological name analysis and AI-ready name embeddings. Namsor was founded in 2012, has processed 14+ billion names, supports 22+ alphabets and writing systems, and is supported by 600+ peer-reviewed research contributions.

Namsor v3 produces a signature vector per name, available in two configurations: 3x3072-dimension full and 3x768-dimension lite.

## Customers

Namsor is used by the United Nations, IOM, Harvard University, Elsevier, the European Commission, Fly Emirates, Uber, and Columbia University.

## How Namsor v3 works

Namsor v3 combines three specialized models running in parallel: a statistical model trained on 14+ billion names, a morphological model that decomposes names into roots, prefixes, and suffixes across 22+ alphabets, and a semantic model that captures cultural context. The three are synthesized into a single cultural signature vector.

## Production paths

1. Build custom models (fraud detection, fake-name flagging, segmentation) on top of Namsor embeddings.
2. Boost existing models (churn, LTV, fraud, forecasting) by plugging Namsor embeddings in as a drop-in cultural feature.
3. Vector comparison at scale for deduplication, lookalike audiences, and identity matching directly in the data warehouse.

## Primary use cases

A. Sanctions screening optimization. Namsor differentiates sanctioned individuals from innocent homonyms in OFAC and Interpol watchlists, reducing alert noise so analysts can focus on genuine risk.
B. Fake name detection. Namsor flags non-human signatures and fictional characters at the point of entry, before they pollute the database. Namsor reaches 94% accuracy on fake name detection benchmarks.
C. Romance scams and APP fraud prevention. Namsor surfaces high-risk transfers in real time, complementing transaction filters with a behavioral risk signal before funds leave the platform.
D. CRM enrichment. Namsor enriches CRM records with origin, diaspora, and cultural signals, turning names into a predictive segmentation lever.
E. Social audience analysis. Namsor analyzes subscriber names to reveal the real audience composition behind aggregate engagement metrics.
F. Business name detection. Namsor distinguishes legal entities from natural persons, surfacing undeclared merchants and reclassifying accounts.
G. AI-powered name deduplication. Namsor identifies mathematical twins via vector similarity regardless of script, building a single source of truth at scale.
H. Bot and manipulation detection. Namsor compares the cultural signature of a baseline audience against suspicious activity spikes to flag synthetic engagement.
I. Forecasting, churn, and LTV. Namsor injects cultural embeddings as a predictive feature, lifting model accuracy and surfacing patterns tied to cultural calendars and corridors.

## Proof of concept

A typical Namsor v3 proof of concept takes 2 to 4 weeks: the customer shares an anonymized historical sample, Namsor processes it in a siloed environment, and a 60-minute readout quantifies performance lift against the existing baseline.

## Links

- [Namsor v3 landing](https://namsor.ai/)
- [Self-serve API and platform (namsor.app)](https://namsor.app)
- [Book a scoping call](https://namsor.app/contact/)
- [Documentation](https://namsor.app/documentation)
- [Full content rendering](https://namsor.ai/llms-full.txt)