Digital White Papers

I L T A W H I T E P A P E R & S U R V E Y R E S U L T S | L I T I G A T I O N A N D P R A C T I C E S U P P O R T & C O R P O R A T E L E G A L D E P A R T M E N T S 30 relevance or the set of documents needing to be evaluated will require building a new model. TAR 1.0 solutions front-load responsive documents during the review process to quickly provide important information to the review team, offering an advantage over traditional linear review. Additionally, the control set used in the TAR training process provides an estimate of the number of responsive documents expected to be found, and once the entire population is scored, it becomes clear how many documents will need to be reviewed in total, allowing for teams to gain efficiencies via workflow planning for review. Criticisms of TAR 1.0 solutions are typically based on the limitations of its one-time training, which does not allow the system to adjust based on information gained during the review. This can be problematic when SMEs are required to code training and control sets prior to synthesizing their own knowledge of the matter. Because one- time training relies on early coding of a training set, the possibility of bias in the predictive model could introduce concerns about the sufficiency of a production. Additionally, low-richness scenarios might delay – if not prevent – the adequate training of a predictive model. Because the full- scale review does not begin until the model is trained, precious time can be lost waiting to see if a predictive model is working. Finally, one-time training is incompatible with rolling collection of documents. Models would need to be retrained to accommodate subsequent loads of data. TAR 2.0 In the second generation of technology-assisted review solutions, TAR 2.0, the underlying technique of continuous active learning (CAL) was specifically adopted to improve upon the challenges that one-time training presented for TAR 1.0. Continuous learning reflects that the predictive model updates throughout the review based on all the coding decisions that humans make, rather than as a discrete step at the beginning of a TAR process. Active indicates that the system uses the updated model to promote the documents with the highest probability of being responsive to the top of the review queue. TAR 2.0 processes can supplement batches of highest-ranked documents to improve efficiency. Examples include documents required to obtain a statistically valid yield estimate to enable workflow planning efficiencies, or a selection of the documents about which the model is least certain, enabling it to gather information needed to make better predictions. TAR 2.0 solutions allow review to begin immediately, without preliminary training. While it may be preferable to have SMEs involved in the early review, this is not a strict requirement as the model will eventually smooth over inconsistent decisions. The low upfront training investment "TAR 2.0 solutions allow review to begin immediately, without preliminary training."

• Cover