Automated Document Profiling - A Predictive Coding (TAR) Engine
Today’s litigators want to move beyond traditional linear review (starting at the first document and progressing to the last). While analytics based on metadata and keyword/concept search and clustering are useful to reduce the data set and prioritize the review, in some circumstances more may be needed. Very large data sets or significant limitations on funding or resources, for example, may require a more sophisticated content-based analysis, to code each un-reviewed document based on the probability that it will be relevant, or privileged, or hot, for example, based on supervised learning methodologies.
Predictive Coding, also known as Technology Assisted Review (TAR) or Computer Assisted Review (CAR) has become an industry standard for advanced content analysis of eDiscovery. The predictive coding process extrapolates human coding decisions made on a sample of the document population across the remaining data, providing a percentage score for responsiveness, privilege, etc. – a prediction of how each un-reviewed document would likely be coded if also reviewed manually. Predictive coding/TAR can be used for several purposes, such as surfacing potentially privileged or “hot” documents for early review by core members of the litigation team. The process can also be used to prioritize the review – pushing the most important documents to the front (those with higher responsiveness scores) for review by higher-paid attorneys, and having lower-paid or contract attorneys review the remaining documents. In certain circumstances, case teams might use predictive coding to support the position that review should stop altogether at a certain point, to avoid the time and expense of reviewing documents with responsive scores below a certain percentage. These strategies address the core goal of lower cost and driving better, faster decisions.
eZAnalytics Engine is the data analysis engine powering the eZSuite modules. The Capital Novus predictive coding/TAR solution (Automated Document Profiling – ADP) offers a convenient and intuitive workflow to code, qualify, prioritize and/or quality assess a litigation review population based on supervised learning methodologies. Integrated into the eZReview feature set, predictive coding is a powerful technology, accepted within the eDiscovery industry for generating statistically validated classifications against a large data universe.
A Content Intelligence and Data Mining Engine