This presents significant due diligence challenges, including:
- The information requirements checklist is much more extensive than for a retail bank.
- Names and their various permutations need to be searched with linguistic and cultural sensitivity.
- There is often a lack of result confidence as identity ambiguity is more prevalent.
- There are additional obligations to determine and corroborate source of wealth, identify associated party risk and support the journey to wealth with the appropriate narrative.
- Multiple sources in multiple languages need to be searched to ensure no risk-relevant intelligence is missed.
When performing adverse media checks on wealth management clients, it would be somewhat foolish to assume your potential Russian or Chinese UHNW client would be written about in the English language or even Latin alphabet alone.
Natural Multi Language Processing technology identifies all textual elements in a piece of text, regardless of source language / script. It reviews findings in native language first to avoid any loss of meaning pre-translation and then produces an English version for review.
Documents like news articles can contain rich information about a person or a company, some of it suggestive of risk, much more of it contextually useful to know. Identifying and organising that material in English, let alone in a foreign language is a demanding task. That’s where our industry- leading NLP comes in.
It goes without saying that if there is an international element to a bank’s client base, whether that’s birth, domicile or footprint, monitoring English language content alone will obviously expose the bank to the risk of missing important intelligence. It stands to reason that extracting intelligence with precision from foreign language sources is a must. However, some approaches machine-translate non-English material into English first and then derive results based on the translation. With such an approach, ‘lost in translation’ doesn’t just mean loss of nuance but potentially and critically, a loss of original meaning.
Take the following two example:
1. 黎明从未成年起就抢劫伤人，二进宫以后仍然不悔改
This Simplified Chinese excerpt is machine translated into English as: “Dawn has never robbed and wounded people since he was an adult.”
Whereas processing in its native language first gives the true English meaning as: “Li Ming has started robbing and hurting people since he was a juvenile, and never repented after he was re-incarcerated .”
2. It’s not just those non-Latin languages where these mistakes can be found. Problems can arise closer to home. “Attaqué en diffamation par Cédric Herrou, Eric Ciotti relaxé.”
Machine translates as: “Defamed by Cédric Herrou, Eric Ciotti released.”
But the actual meaning is the opposite: “Accused of defamation by Cédric Herrou, Eric Ciotti is released.”
And isn’t it ironic that in our KYC / AML world, one of the most familiar words, ‘sanction’, makes the case for a semantic approach as it can mean opposing things, depending on whether it is a noun or a verb?
These apparently minor mistakes could have major consequences. Stripping away the syntactical, grammatical and lexical rules of the original language of the source material for the sake of the ‘convenience’ of processing everything in English will prove an expensive shortcut.
It is prudent for wealth management firms’ KYC screening processes to operate as effectively in most Indo-European languages and the main non-Latin scripts (e.g. Russian, Arabic, Chinese, Hindi, Thai).
In line with market trends, smartKYC’s software now also analyses Hindi in its original script using Natural Language Processing. This is particularly pertinent for wealth managers as recent research from Knight Frank suggests that India will see a 63% increase in ultra-high net worth individuals by 2025. This will outpace the global average of 24% and the Asia average of 38%.
With such intelligent automation, private banks and wealth managers can empower their front-of-house teams to be an efficient and effective first-line of KYC defence and focus their compliance teams where it really matters.
