Rapid and efficient case opening from negative news
Disclosed is an approach in which news alerts are scanned in real-time or near real-time, relevant alerts identified through a topic extraction model, and associated actors identified through an entity extraction model. An entity resolution model may be applied to determine which actors are clients. The topic extraction, entity extraction, and/or entity resolution models may apply, for example, natural language processing models. The alert may be enriched by being packaged with client and transactional data to generate an enriched alert. A predictive model may be applied to the enriched alert to identify events with a high probability of law enforcement referral, and the enriched alert may be automatically transmitted to certain identified devices. The predictive model is trained using a combination of news alerts and data on clients and transactions, yielding enhanced predictions. The disclosed approach reveals relevant underlying associations and connections between entities and transactions not otherwise feasibly discoverable.
1. A machine-learning method comprising:
monitoring, by a computing system, one or more news sources, which generate news alerts, to detect a set of news alerts satisfying a criterion;
applying, by the computing system, an entity extraction model to the set of news alerts to identify a set of persons or organizations associated with the criterion;
generating, by the computing system, an enriched alert comprising one or more of the news alerts in the set of news alerts and data on the set of persons or organizations;
applying, by the computing system, a predictive model to determine a probability that the enriched alert will trigger a suspicious activity report;
determining, by the computing system, that the probability is greater than a threshold probability;
applying, by the computing system and in response to determining that the probability is greater than the threshold, a relevance model to identify a destination computing device; and
transmitting, by the computing system, the enriched alert to the identified destination computing device;
wherein applying the entity extraction model produces a first person or organization named in the set of news alerts, and a second person or organization who is not named in the set of news alerts but who is related to the first person or organization; and
wherein the news alerts include at least one of written alerts or audiovisual alerts.
2. The method of claim 1 , wherein monitoring the one or more news sources comprises monitoring the one or more news sources in real time.
3. The method of claim 2 , wherein the one or more news sources are monitored by accessing multiple third-party devices.
4. The method of claim 1 , wherein the computing system is associated with an entity, and wherein the method further comprises determining, by the computing system, that at least one of the first person or organization or the second person or organization is a client of the entity.
5. The method of claim 1 , wherein the first person or organization and the second person or organization are related by having engaged in one or more financial transactions with each other.
6. The method of claim 1 , wherein the first person or organization and the second person or organization are related by being in a same family.
7. The method of claim 1 , wherein the first person or organization and the second person or organization are related by being involved in a same commercial enterprise.
8. The method of claim 1 , wherein generating the enriched alert comprises retrieving, by the computing system, the data on the set of persons or organizations from one or more internal databases.
9. The method of claim 1 , wherein the criterion is an apparent commission of a crime.
10. A machine-learning method comprising:
monitoring, by a computing system, one or more news sources, which generate a set of news alerts, to detect a set of news alerts satisfying a criterion;
applying, by the computing system, an entity extraction model to the set of news alerts to identify a set of persons or organizations associated with the criterion;
generating, by the computing system, an enriched alert comprising one or more of the news alerts in the set of news alerts and data on the set of persons or organizations;
applying, by the computing system, a predictive model to determine a probability that the enriched alert will trigger a suspicious activity report;
determining, by the computing system, that the probability is greater than a threshold probability;
applying, by the computing system and in response to determining that the probability is greater than the threshold, a relevance model to identify a destination computing device;
transmitting, by the computing system, the enriched alert to the identified destination computing device; and
determining, by the computing system, that one or more of the persons or organizations in the set of persons or organizations is a client of an entity associated with the computing system, wherein the data on the set of persons or organizations in the enriched alert was known to the entity prior to the identifying the set of persons or organizations associated with the criterion;
wherein generating the enriched alert comprises retrieving, by the computing system, the data on the set of persons or organizations from one or more internal databases.
11. The method of claim 10 , wherein the criterion is apparent commission of a crime.
12. A computing system comprising one or more processors configured to:
monitor one or more news sources, which generate news alerts, to detect a set of news alerts satisfying a criterion, wherein both audiovisual news alerts and written news alerts are monitored;
apply an entity extraction model to the set of news alerts to identify a set of persons or organizations associated with the criterion;
generate an enriched alert comprising one or more news alerts in the set of news alerts and data on the set of persons or organizations;
apply a predictive model to determine a probability that the enriched alert will trigger a suspicious activity report;
determine that the probability is greater than a threshold probability;
apply, in response to determining that the probability is greater than the threshold, a relevance model to identify a destination computing device; and
transmit the enriched alert to the identified destination computing device;
wherein the computing system, by applying the entity extraction model, is configured to produce a first person or organization named in the set of news alerts, and a second person or organization who is not named in the set of news alerts but who is related to the first person or organization.
13. The system of claim 12 , wherein the computing system is configured to access multiple third-party devices to monitor the one or more news sources in real time.
14. The system of claim 12 , wherein the second person or organization is a client of an entity associated with the computing system.
15. The system of claim 12 , wherein the computing system is configured to detect the set of news alerts by applying natural language processing models to the set of news alerts.
16. The system of claim 12 , wherein the computing system is further configured to determine that one or more of the persons or organizations in the set of persons or organizations is a client of an entity associated with the computing system, wherein the data on the set of persons or organizations is retrieved from one or more internal databases of the computing system.