Originally named predictive coding, this process has since been renamed TAR (Technology Assisted Review). The original TAR methodology has been improved upon with TAR 2.0 and, most recently, TAR 3.0. All of these TARs have one thing in common – they require a subject matter expert, or SME, (typically a senior lawyer or client representative) to train the computer system. Until recently, TAR involved using a Latent Symantec Indexing system (commonly called a conceptual engine) to identify documents that contain topically similar content – if one document discussed oil wells, and another discussed transporting oil from a well to a refinery, they were identified as discussing similar topics. Conceptual systems generally work well. Sure, in some cases, they fail to find records a knowledgeable lawyer would clearly identify as relevant, but generally, the technology has proven to be accurate enough. However, TAR can be so much better.
Back before predictive coding arrived on the scene, well developed entity extraction technology was already available. Entity extraction analyses sentence structure to identify people’s names, places, dates, financial transactions, and brand names, among other things. An entity analysis will determine that someone with a last name “Rock” is not the same as a big stone or a type of music (concept systems do this too, but not as well). One could conceptually search for the term counterfeit and see that the concept engine also suggested knock-off as a similar term. However, since “knock-off” could also refer to the time to stop working (as in “What time do you knock off?”), using the results of the concept search alone would bring in false hits. By following up the concept search with an entity search that would look for the terms “counterfeit” and “knock-off” where they are associated with brand names, such as “Gucci”, the results would be much more focused and accurate. The smoking emails that contain colloquiums, or slang, are less likely to be missed.
Understanding the technology, and how to combine different techniques to arrive at a more accurate result, is a different skill set than the legal expertise of the SME. To obtain the results when adopting TAR, working side by side the SME should be a professional with the technical expertise to assess the tools and the process. The combination of both skill sets gives the best results.