Real-Life Robocop: How NLP Is Fighting Financial Crime

“There used to be a point where chatbots could barely understand slang. Now, they’re able to recognize numerous languages and styles.” The amazing progress artificial intelligence has made in human language processing fascinates Kfir Bar, chief scientist at Basis Technology. For years, Bar has explored the gambit of natural language processing disciplines and challenges. Bar started this journey at Tel Aviv University where he received his Ph.D. in NLP focusing on machine translation.

[Related Article: How Security Agencies Use AI to Stop Crime]

“Some languages, like Arabic, introduce additional challenges,” he said. “Its morphology is much richer than English. In Arabic, you can modify one word in hundreds of different ways to express different syntactic and semantic roles.”

The ever-advancing NLP techniques that Bar and other researchers are working on affect our daily lives in considerable ways. Bar discussed how people in his field are using the developing technology to advance security and anti-laundering/financial crime tools.

How do we apply NLP to security?

Threats made online can be analyzed regardless of the language they’re written in. Bar said by analyzing texts in a post—such as those on social media platforms—AI can discover explicit and implicit intent of bad actors.

“Security is about prevention,” he said. “If you are able to understand ahead of time that someone has the willingness of doing something bad, then you can try to prevent that.”

Understanding intent is challenging: It essentially means AI must know what someone believes in based on what they write and predict their actions accordingly. This requires a thorough understanding of the relevant context, including the culture, ideology, sentiment, and emotional state of the authors of a given post, Bar explains.  

How is NLP being used to fight financial crime?

The increasing prevalence of cross-country money transactions and increasing ingenuity of financial-crime techniques create serious challenges for the financial institutions tasked with identifying potential and actual criminals and criminal activity.

According to a 2017 report from the National Center for Victims of Crime, the number of financial crime complaints reported to the Federal Trade Commission is rising. There were more than 3 million complaints of fraud, identity theft, and other financial crimes reported to the FTC in 2015—up from less than 1 million a decade earlier.

But AI can help, and know your customer (KYC) systems are a perfect example.  

FIs are expected to not only vet prospects before they become customers but periodically during the entire life of the business relationship. To do this, FIs have to perform background checks that involve a variety of information sources, often including international watchlists, news articles, social media mentions, and proprietary databases.

“The goal of this process is to find any relationship between the potential or actual client and bad actors, dangerous organizations, and/or negative behaviors,” Bar said. Considering the number of prospects and clients FIs handle and the amount of relevant information that would need to be processed to answer these questions in a meaningful way, automation is really the only answer.

In contrast to human systems, NLP applications can process information quantities at internet scale. They can also be trained to find and extract vital intelligence—the people, organizations, places, and their inter-relationships—buried in these information sources.

With such technology under the hood, KYC systems can live up to the demanding expectations FIs are held to.

[Related Article: Detecting Cybersecurity Incidents with Machine Learning]

Conclusion

While there are applications across every industry for this technology, NLP’s ability to help in the fight against financial crime is much of what Bar and his colleagues focus on. Healthy societies rely on functioning financial systems, and he believes this is something believe NLP can effectively support.  

“NLP excels at the automated analysis of huge quantities of unstructured data, so it’s a powerful resource for financial institutions as they combat fraud, money laundering, and criminal enterprise generally,” Kfir notes. “However, it’s not the only tool. NLP is one of a number of technologies—each handling different types of data—that should be used together. The more this happens, the safer our financial system will be.”