Our Insights

Thought Leadership and Industry Trends

Home 9 Insights 9 Automation 9 Automating eDiscovery with Active Learning AutoRanking

Automating eDiscovery with Active Learning AutoRanking

Nov 3, 2022

The greatest challenge in any eDiscovery matter is to capture the entire universe of documents that may be relevant and then cull non-relevant documents as quickly as possible to expedite review and production. The standard method of culling involves a tedious process of search term trial and error followed by review to identify the relevant population and weed out the non-relevant document set. Even with the most precise search terms and filters, the majority of the population may still be composed of non-responsive documents.

While many case teams have adopted Technology Assisted Review as standard practice on large complex matters, the proper defensible deployment of a TAR workflow requires a high level of commitment and subject matter expertise. Litigators and attorney technologists must work in unison to establish defensible workflows and protocols with statistically valid sampling, confidence level and elusion rates to assuredly defend TAR workflow against potential challenges by opposing counsel.

These considerations often dissuade legal practitioners from leveraging all of the powerful eDiscovery tools and technology at their disposal, favoring the “safer,” more time-intensive and costly route of standard linear review. It is possible, however, to strike a beneficial balance between a full blown TAR project and standard linear review by leveraging Relativity Active Learning .

Active learning can be a powerful tool to pull the most relevant documents to the surface, like a magnet, expediting review by automatically prioritizing the most important documents first. Machine learning analyzes coding decisions on a rolling basis to build models and rank all documents in the project, pushing the most highly ranked, potentially relevant documents to the front of the review queue. But what happens if a case team is required to review and produce priority documents in a particular order, such as by custodian, date range, or some other criteria?

Vision AutoRank allows clients to fully realize the benefits of Relativity Active Learning at any stage of the search, review, and production process without the need to follow prioritized batching. Dashboard visualizations, searchable fields, and coding layouts will be automatically updated to display the likelihood (score) predicting how a document or set of documents will be tagged based on up-to-date machine learning from prior tagging on documents with similar characteristics. AutoRank helps expedite review, improve search results, and promote greater coding accuracy and consistency, while also maintaining the flexibility to review documents based on matter-specific production priorities.

To read the next blog in the series, click here.

About the Author

<a href="https://cdslegal.com/team/michael-milicevic-esq/" target="_blank">Michael Milicevic, Esq.</a>

Michael Milicevic, Esq.

As VP eDiscovery Products & Solutions, Michael Milicevic leads the Global CDS Product Solutions Team, designing and developing proprietary products and solutions and expanding the CDS technology portfolio. He also oversees all Midwest operations. He has over 12 years of eDiscovery and legal experience and has worked on multiple large matters for a variety of clients, including Am Law 100 firms, global financial services organizations, and multi-national manufacturers.