Entity Extraction for Adverse Media Monitoring Strengthens AML

Entity Extraction, Risk Management, Social Media Analysis

Entity Extraction for Adverse Media Monitoring for AML

Financial Institutions Have KYC Obligations

Financial institutions need to avoid doing business with questionable entities engaged in money laundering activities so as not to support criminals’ illegal financing unwittingly. These institutions also would not want to risk their reputation or pay heavy fines for violations.  To avoid this unfortunate outcome, they are obliged to establish a Know Your Customer (KYC) process.

KYC refers to due diligence performed by financial institutions, and it typically occurs when a customer is first on-boarded in order to verify their identity and assess the potential risks of dealing with it. KYC is a complex process – it starts with verifying basic facts about an entity, such as address of its headquarters and its true ownership. It includes verification of the sources of funds, as well as checking sanctions lists. And of course, this due process needs to be done periodically.

Adverse Information Contained in the World’s Media Is Increasingly Important to Vetting Customers

A critical part of this process is monitoring adverse media. Adverse media is any type of unfavorable information coming from a very broad variety of news sources. These include traditional news outlets and other sources like blogs or social media.

Checking individuals and companies against these adverse media sources may reveal entanglement in a broad array of criminal activities such as:

  • Money laundering
  • Drug or human trafficking
  • Bribery
  • Organized crime
  • etc.

These activities linked to potential customers pose a serious threat to a firm’s reputation and will certainly raise the risk calculation made by a financial institution.

Current Collection of Adverse Information Is Too Slow and Incomplete

The typical current approach to checking media sources involves a staff of AML analysts reviewing daily hundreds of news articles and other unstructured data sources (which typically are produced by a third party data provider). They are simply looking for adverse information about potential clients. In doing this, the analysts also need to determine if the information truly is about the entity they’re researching. This approach is slow and not very efficient. It also has no chance of scaling to the extremely large amount of media available in today’s digital world. There is a technology that can help, though.

Advanced Entity Extraction for Adverse Media Monitoring

Entity Extraction (aka Named Entity Recognition or Named Entity Extraction) is a technology that automatically identifies key semantic concepts in unstructured text data, such as people, organizations, and places. It can do this at scale, handling extremely large amounts of digital data.

Most extraction products focus on names. This basic named entity extraction is necessary but not sufficient for monitoring adverse information. We need to go beyond that. There are two important advanced entity extraction capabilities that are necessary for adverse media monitoring:

  • Relationship Extraction

Relationship Extraction finds all the links that an entity may have with other entities: for instance, it uncovers the fact that a principal of a company was previously employed by another company that was involved in insider trading.

  • Event Extraction

Event Extraction finds events containing adverse information that individuals or corporations have participated in. Event Extraction finds not only an adverse event like “indict” but also the participants in the event (e.g., who was indicted). It performs a complex natural language processing (NLP) analysis. One of the main advantages of NLP-based Event Extraction, compared to simpler proximity-based approaches, is that it understands the syntax and semantics of unstructured text, so it can identify the specific roles of individuals in events, i.e., whether they were the ones issuing the indictment or the ones indicted. This capability is critical in utilizing adverse media monitoring for KYC. In effect, Event Extraction reduces the number of false positives instances where the person of interest is recognized but incorrectly associated with an event of interest.


Adverse Media Monitoring supported by Entity Extraction is a critical tool in Anti-Money Laundering. It allows financial institutions to get a very detailed and accurate view of potential clients and any baggage they may be carrying.