Continuous active learning: the EDT approach

The number of items involved in regulatory and legal investigations continues to grow, enormously. This places an increasing burden on those responsible for reviewing and evaluating large collections of digital evidence. Continuous active learning (CAL), the second generation of machine learning technologies for document review, has great potential to accelerate investigations and reduce the burden of linear review.

This paper examines the theory, judicial rulings, and practice of CAL and shows how EDT’s implementation learns from each classification made by human experts to improve its classification predictions with every review batch.