Links

Text Matching AI

Why Do We Need Text Matching AI?

Name, the most important value for verifying one's identity, has an aspect that one can be expressed in different ways as shown below.
This characteristic of name cause unexpected problems throughout the KYC and AML processes.
One of them is when you have to judge 'the two are the same person?', at point of AML screening and duplicate approval prevention.
For example, if a exchanger carries out a promotion for KYC verification, the best scenario for them will be 'One verification for one'.
But what if a user proceeds first KYC with local Id and second KYC with international passport, how do we figure out that it's actually one person?
To solve this problem, ARGOS adopted 'Text Matching AI' to compare names.

What is Text Matching AI?

Text matching AI is an algorithm that scores the similarity of two text. We try to deal with typing errors, nicknames, initials, titles with it. The AI translates 14 languages and applies 15 matching logic.

Supporting Languages

Arabic
Burmese
Chinese, Simplified
German
English
French
Chinese, Traditional
Greek
Hebrew
Hungarian
Italian
Japanese
Khmer
Korean
Pashto
Persian
Portuguese
Russian
Spanish
Thai
Urdu
Vietnamese

Example of Matching Logic

Phonetic similarity
Jesus ↔ Heyzeus ↔ Haezoos
Transliterate spelling differences
Abdul Rasheed ↔ Abd al-Rashid
Nicknames
William ↔ Will ↔ Bill ↔ Billy
Missing spaces or hyphens
MaryEllen ↔ Mary Ellen ↔ Mary-Ellen
Titles and honorifics
Dr. ↔ Mr. ↔ Ph.D.
Gender
Jon Smith ↔ John Smith (but not Joan Smith)
Truncated name components
McDonalds ↔ McDonald ↔ McD
Missing name components
Phillip Charles Carr ↔ Phillip Carr
Out-of-order name components
Diaz, Carlos Alfonzo ↔ Carlos Alfonzo Diaz
Initials
J. E. Smith ↔ James Earl Smith
Names split inconsistently across database fields
Dick Van Dyke ↔ Dick Van . Dyke
Same name in multiple languages
Mao Zedong ↔ Мао Цзэдун ↔ 毛泽东 ↔ 毛澤東
Semantically similar names
Eagle Pharmaceuticals, Inc. ↔ Eagle Drugs, Co.
Semantically similar names across languages
Nippon Telegraph and Telephone Corporation ↔ 日本電信電話株式会社
Organizational Aliases
IBM ↔ Big Blue

ARGOS KYC with Text Matching AI

With years of KYC know-how, ARGOS has found a way to better clarify names. It enables us to offer two values to our clients.
First is automated KYC. Even when the result from OCR AI differs form user input due to misspell etc, you will get the KYC status result immediately without waiting for manual review.
Second one is to filter out the duplicated users which becomes quite significant issue for verifying users. Even when the type of ID changes or names are in different languages, we will catch them.