Responding to an Anti-Trust DOJ Second Request

Using Reveal Data when responding to an anti-trust DOJ second request: Global law firm McDermott Will & Emery (MWE), were required to produce documents to the Department of Justice for an anti-trust Second Request. They were against tight deadlines, with very high stakes.

Using Reveal’s NexLP technology and clever consulting, they were able to produce the documents on time and eliminate the need for exhaustive manual review.

**A second request is a discovery procedure by which the Federal Trade Commission (FTC) and the Anti-trust Division of the Justice Department (DOJ) investigate mergers and acquisitions which may have anticompetitive consequences (United States anti-trust law).

Challenges: DOJ Second Request

Data size and time constraints played a huge role in how the review was structured. The original data collection started with approximately 1.3 million documents. This, coupled with time constraints meant that "traditional" machine learning training would take too long (10-14 days).

Using Reveal Data to Respond to DOJ Second Request

Reveal’s NexLP technology, a cognitive analytics tool, was leveraged to respond to the request in a very short timeframe. This technique eliminated the need for exhaustive manual review.

NexLP AI uses Machine Learning and AI to mine the data for patterns and anomalies, map custodian relationships, conversations, and documents. The superior technology does so in a fraction of the time other applications require. Reveal’s NexLP technology more closely represents human intellect and helps find the most important documents as quickly as possible.

“The combination of conversation mapping, sentiment analysis, and the Continuous Active Learning engine is a powerful tool that allows our teams to identify the most important information in a data set while defensibly eliminating irrelevant documents"
Martha Louks, Director of Technology Services at McDermott Will & Emery

Outcome

Guided by Reveal's NexLP Advanced Machine Learning, the MWE team completed the entire project in 37 days. This included total time from collections through to production. Subject matter experts trained the system in 3.5 days. This was much faster than the typical 10-14 days required in “traditional” machine learning coding. The system reached stabilisation with only 1.3k sample documents reviewed. This was also much faster than review of the usual 8k – 10k documents typically necessary in traditional machine learning applications.