Contact us and see what NetOwl can do for you!
Entity Extraction Helps Companies Know Who They are Dealing with
Companies Have to Stay Clear of Corrupt Activities in Foreign Countries
Onboarding a partner company or an executive with a history of illicit activities exposes a company to great risk including being unknowingly involved in criminal activity. Substantial damage may be done to a company’s reputation, and its business may be badly affected. This is particularly true of international businesses that may have a foreign subsidiary or partner engaging in corrupt practices. An American law, the Foreign Corrupt Practices Act (FCPA) of 1977, imposes substantial penalties for activities such as bribing foreign officials to facilitate business. There have been many instances of this.
Penalties for a Misstep Are Substantial
The subsidiary or partner on the ground may well think it’s fine to engage in such activities. They may even think it’s necessary to play by the local rules to win business. If discovered, though, the whole firm may be subject to substantial financial and criminal penalties. American companies even sometimes hire what’s called a due diligence firm to vet overseas intermediaries and to make sure they are not dealing with foreign government officials embedded in an otherwise privately held foreign company.
It’s therefore very important for companies to find all adverse information regarding companies and staff that would increase the potential risk of damage to the firm’s reputation. For instance, a company needs to know whether a candidate for an executive position may have past indictments, arrests, or prosecutions, or whether a prospective partner firm may have been sanctioned or may have even had criminal activity in their past. Increasingly a necessary part of this is checking out any reports about their activities on traditional news media outlets as well as on social media.
Monitoring the Media Using AI-Based Entity Extraction
One new way of performing this adverse media monitoring is to use AI-based Entity Extraction technology (aka Named Entity Extraction or Named Entity Recognition). Clearly, given ever-growing volumes of data, monitoring news media on an ongoing basis, even if only in English, is beyond purely human capabilities. Social media is even more of a challenge.
Entity Extraction Understands the Contents of Unstructured Text
Entity Extraction automatically identifies key semantic concepts in unstructured text data, such as people, organizations, and places. Yet, although extracting these basic concepts is necessary and important for monitoring adverse information, it’s necessary to go beyond this level and extract relationships and events.
Relationship and Event Extraction Finds the Adverse Information
Relationship Extraction finds the associations that an entity may have, such as a person’s employment affiliation. As you can imagine, if a potential executive hire has been affiliated with a sanctioned organization or government, that would be something a prospective employer would want to be aware of.
Event Extraction finds events conveying adverse information and the entities (e.g., individuals, corporations) that participated in them. For instance, given a sentence like “Michael Avenatti tried to extort $25 million from Nike “, Event Extraction doesn’t just discover that a key word like “extort” occurs in close proximity to “Michael Avenatti” and “Nike”. It performs linguistic analysis and is able to recognize that “Michael Avenatti” was the perpetrator of this criminal event while “Nike” was the victim. Furthermore, Event Extraction can detect when an ambiguous word like “charge” is used in an adverse vs. non-adverse sense. For instance, Event Extraction can tell that “Steve DiPietro was charged with tax fraud” is adverse information about DiPietro, whereas “Steve DiPietro was charged with investigating tax fraud” is not.
By analyzing the syntax and semantics of unstructured text, Event Extraction can more accurately identify adverse events and the specific roles that various entities play (e.g., perpetrator vs. victim), thus extracting only relevant events and identifying the bad actors. In technical terms, Event Extraction reduces the number of false positives, that is, instances where the entity of interest is recognized but is not a bad actor.
In sum, Entity Extraction together with its more advanced capabilities Relationship Extraction and Event Extraction are necessary tools for an organization’s risk and reputation management.