Updated: Feb 14, 2020
Many attorneys use technology assisted review (“TAR”) solutions to help streamline their practice. It can help with document review, case assessment, contracts, and other litigation tasks. One major draw of TAR is that it reduces effort and cost related to eDiscovery, while also improving accuracy. Numerous studies have concluded that TAR provides results that are superior to manual review. As the years go on, this software continues to improve and provide even more benefits to the legal industry.
What is Technology-Assisted Review (TAR) 1.0?
The earlier version is referred to as TAR 1.0. This includes programs that incorporate simple passive learning (“SPL”) and simple active learning (“SAL”). Either a human operator (SPL) or computer (SAL) will select documents for human review and coding. A knowledgeable reviewer will code the documents for relevancy so they can be used as training examples. Training will need to be repeated until the system is stable, which is when it no longer gets better at identifying relevant documents in the control set. The software then builds a classification/ranking algorithm that will pull in other relevant documents.
What is Technology-Assisted Review (TAR) 2.0?
TAR 2.0, often referred to as continuous active learning (“CAL”) is a refined method that contains several upgraded features. Often characterized as “supervised machine learning,” the software uses a search engine to run a simple query of a few terms against a document collection. The search engine uses relevance ranking to present the reviewer with documents that are likely relevant to the inquiry. If the results are accurate, these documents may be used as the training set. The program continues to choose the documents that are most likely relevant, which are then reviewed, coded, and used to improve the system. This occurs until it can no longer find any more relevant documents.
Three Reasons Why Lawyers Should Upgrade to Continuous Active Learning (CAL) Technology
TAR 2.0 is an advanced technology that saves organizations even more time and money than before, while providing the most efficient results on the market. TAR 2.0 is worth a try because it remedies the following limitations from its previous version:
TAR 1.0 is time consuming: While quicker than manual review, TAR 1.0 requires senior attorneys, who bill at high rates, to spend significant time looking at marginal documents to train the system. The reviewer will oftentimes have to review the same documents more than one time. TAR 2.0 technology requires significantly less review efforts because the computer can search for relevant documents. There is also generally no second review, because by the time the computer stops learning, all documents it deemed relevant have already been identified and manually reviewed.
TAR 1.0 is random: Because TAR 1.0 gives the reviewing attorney a sample that is completely random, there is no assurance this group is in fact representative of the larger population. TAR 2.0 provides a more efficient system where the software can identify relevant documents and provide the reviewing attorney with a smaller and more refined set of documents to review. Because of the superior algorithm, the results are far more accurate.
TAR 1.0 generally can only work off one data set: Adding new documents to the system can render the control set invalid, as they were not part of the random selection process. TAR 2.0 software fixes this because it is programmed to review and code documents on a rolling basis. This is more practical in a litigation setting, where new documents are being discovered and disclosed throughout the case.
Based on these features and upgrades, there is no question that attorneys should explore TAR 2.0 solutions. As the supervised machine learning involved in CAL becomes more autonomous in training and execution, TAR 2.0 will continue to improve and provide more benefits, saving every party involved even more time and money.