Matthias Dennig/ July 30, 2018/ Uncategorized

From typo to terrorist financing? Applied linguistics in compliance

When the EU imposed sanctions on the Gaddafi regime in early 2011, no one could have foreseen the ensuing problems for compliance. A lack of clear standards for the spelling of Arabic names in Roman script gave rise to dozens of variations of the Libyan leader’s name. From Muammar Qaddafi and Muammar Al-Gaddafi to Elkaddafi, very few of these featured in the sanctions lists. For the EU, the responsibility for ensuring all possible spellings were checked, lay firmly in the hands of the financial institutions. Reason enough for compliance to get to grips with the transliteration/transcription issue.

Transliteration vs. transcription

In transliteration, individual letters or characters from one alphabet or writing system are transposed into corresponding characters from a different alphabet. For example, to convert a Greek, Cyrillic, or Arabic word into Roman script, a suitable letter from the Roman alphabet must be found for each character. This one-to-one transposition makes the process fully reversible. If the source script has more characters than the target script, diacritical marks are used. Transliteration is language-neutral. This means that the spelling of a word is not aligned with a specific target language – and the way that word is written bears no direct correlation to its pronunciation.

In transcription, on the other hand, the aim is to find a way of writing a word that encourages its correct pronunciation in the target language. In this way, differing rules of pronunciation in German and English, for example, result in words being written differently, even though both languages use the Roman alphabet. The spelling of a word is therefore less likely to point clearly to its written form in the original script. In transcription, not everyone follows the established guidelines to the letter (e.g. ISO standards). This has resulted in numerous variations and dozens of spellings of one and the same name.

قطر (Arabic original)
qṭr (pure transliteration)
Qatar (English transcription) [1]

Transliteration and transcription against terrorist financing, money laundering, and embargo breeches

If no solution can be found to identify spelling variations, the risk of falling prey to fraudsters will always be there. Just a single letter can stand between compliance and non-compliance. This means not only checking the names of individuals and companies, but also verifying enterprises’ legal status (e.g. foundation or charity) and any official suffixes (e.g. Inc. or Ltd.). An unequivocal identification of customers is also dependent on reviewing relevant information in lists of sanctions and politically exposed persons (PEPs).

Within the EU, there is already demand for transliteration and transcription. Examples include:

  • The use of umlauts and special characters in German
  • Customers who come from countries with different scripts e.g. Cyrillic, Arabic, Chinese

Thorough screening and cross-checks are particularly critical in the case of new clients (customer onboarding CDD, KYC), but must also be part and parcel of continuous monitoring (ongoing CDD, payment screening). Although many payments are still processed exclusively in ASCII format, they need to be checked definitely against all transliterated original scripting names on sanction lists. New names, companies, or banks can appear on lists of sanctions and PEPs at any time.

But transcription is not the end of the story. The next step is the development of algorithms that can carry out similarity searches. Essentially this involves taking an English transcription of a name and finding every possible variation.

What next?

Lists of sanctions and PEPs are generally in English and feature one or more written forms of a name – but critically not all. Transliteration and transcription favor a multi-pronged approach:

  1. For each individual new client, vendor etc., all possible variations of a name are researched.
  2. Some official sanctions lists include names in original script. These are also transliterated and all possible variations documented.
  3. In some cases, there may still be lists written entirely in original script. These also need to be transliterated.
  4. Using similarity search algorithms, the variations resulting from all of the above are cross-checked against one another.
Transliteration and transcription: An integral part of compliance

Functionality for transliteration, transcription, and similarity searching has clearly earned its rightful place in a comprehensive compliance system – both tightly integrated and fully automated. Whether new client or ongoing due diligence, hunting down all possible spellings of an individual’s name, a company name, or an enterprise’s legal form will always be a challenge – not to mention cross-checking these variations against each other and current sanctions lists. Only in this way, however, do we stand a chance of preventing an unusual spelling from becoming a loophole for fraud, money laundering, or terrorist financing.


Matthias Dennig

About Matthias Dennig

Matthias Dennig works as a presales and consulting expert for the product SMARAGD aces360. His area of responsibility includes advising prospective clients and supporting the introduction of the SAP BIS based product SMARAGD aces360. For more than 15 years he has worked as a project manager in the compliance environment for targens GmbH in various major projects in Germany and other countries.