Fuzzy name matching for AML/KYC compliance is very challenging due to a variety of inherent difficulties such as typos, spelling differences, aliases and, in some cases, transliteration problems. Chinese names can be particularly difficult, especially when dealing with following two scenarios:

• Matching transliterated Chinese names against real Chinese names

• Matching real Chinese names against real Chinese Names

IMTF’s latest update to its proprietary name screening solution scores precision and recall levels of up to 99.99%. This means: less than one false positive and one false negative per 10’000 records. All this while achieving average speeds of 200-300 ms per record.

In order to achieve these advanced scores, different methods and models need to be applied for specific situations. IMTF has developed a specific process for chinese names, which inlcudes:

1. Deep learning

2. Original Chinese character recognition using the Four-Corner Method

3. Transliteration into various romanized forms using:

   • the detection of accented forms

   • the detection of vowels phonetic similarity

   • other fuzzy scenarios

4. Extensive feature engineering

5. Detection of transliterated Chinese names

➡️ Want to know more about our name screening solution? Contact us, we’ll be happy to provide additional information or arrange a demo.