SyntelliRead: on the spot – Michael Jeske
It is sometimes said that a computer system is only as good as the data with which it operates. Bearing this in mind, human-generated text found in the likes of emails, chat and social networking platforms, which is free-formatted and unpredictable in structure and format, has long represented a human-to-machine translation quandary.
Rather than deploy manual data extraction from the text - which can be slow, costly and error-prone - the financial services industry is turning to innovative text interpretation software. This has become particularly important in the field of anti-money-laundering (AML) systems, which are required to identify transactions in non-cooperative countries and territories, high-intensity drug-trafficking and financial crime areas, and other high-risk geographies.
"In the AML space, financial institutions are discovering that they have serious data-quality issues adversely affecting their systems' ability to assess the geographic risk inherent in their transactions," says Michael Jeske, president at SyntelliRead. "As a result, banks need to acquire third-party software to help their AML systems make sense of the free-formatted party addresses in their transaction data."
SyntelliRead's natural language processing
Focused exclusively on developing natural language processing software for the financial industry, SyntelliRead's flagship product is SyntelliRead® Location Extractor™. Compatible with all AML systems, it accurately extracts and standardises city, province and country names from free-formatted addresses, allowing banks to analyse their transactions geographically and dramatically improving their ability to identify activity in high-risk jurisdictions.
"Address interpretation is a complex problem to solve." says Jeske. "The names of people, organisations and streets can contain city and country names, so the software needs to determine the context in which a name is being used to prevent false positive matches. When an address is incomplete, the software needs to infer the missing geographic information. Simplistic keyword matching technologies lack the functionality necessary to achieve acceptable accuracy rates."
While demand for products such as SyntelliRead Location Extractor is on the rise, Jeske feels banks are still not taking full advantage of new and emerging technologies that are able to mitigate and tackle money laundering and fraud. This, he claims, is in part due to banks being overwhelmed by ever-changing regulatory requirements.
"When banks are consumed with policies, procedures and training issues - as well as handling rafts of regulations coming from numerous bodies - finding and acquiring technology to deal with poor data quality becomes a low priority," he explains.
As the Financial Action Task Force looks to further develop and amend its policies on how to successfully tackle money laundering and terrorism financing, criminals are responding by refining their activities and behaviour, placing an even greater pressure on banks to enhance their AML systems accordingly.
"There is a constant race going on between criminals inventing new and illegal ways to launder money, and banks figuring out how to detect these schemes," says Jeske. "It's a vicious cycle. Financial institutions need to rapidly react to these changes, and thus the technology they acquire has to be sufficiently flexible, configurable and adaptable."
With the endless unpredictabilities associated with free-formatted text, technology players such as SyntelliRead will continue to develop and refine the algorithms found in their software. According to Jeske, given the current scale of money laundering, this will have to be done with a global perspective in mind.
"Institutions can only analyse their own transactions" he says. "It would be advantageous for monitoring to be performed on a more global basis to facilitate the discovery of illegal activity that is being dispersed across many institutions and/or countries; for example, the majority of large value payments are executed via SWIFT messages.
Therefore, we should explore the potential to perform centralised monitoring of activity occurring across the entire SWIFT network."