Exploring Technology Assisted Review (TAR)

Technology Assisted Review is a crucial tool within eDiscovery due to the increasing amounts of data generated and the disproportionate amount of time and costs to carry out an electronic document review.

What exactly does Technology Assisted Review entail?

Technology Assisted Review (or TAR for short) uses mathematical algorithms and statistical sampling to automatically code documents. The software undergoes training using a seed set of documents, all coded by an expert. This aids in  determining what constitutes a ‘relevant’ document and what does not.

How can CYFOR Legal help?

At CYFOR Legal, we use Relativity Assisted Review (RAR) for predictive coding as one of our tools. RAR uses a functionality called ‘Categorisation’ to arrange the documents into groups of ‘Relevant’ and ‘Not Relevant’ documents. Categorisation uses Relativity’s analytics engine to look at textual concepts within a document set. This is based on a type of textual analytics called Latent Semantic Indexing (LSI). The analytics engine will examine concepts within a document and identify other documents with similar textual content. We seize this opportunity to teach the system about the types of documents we are interested in. This allows the analytics engine to categorise them accordingly.

The best fit for you.

CYFOR Legal collaborates with clients in the early stages of a litigation case. This allows us to assess whether TAR is the most suitable approach for the project at hand. Additionally, once a client instructs our consultants to carry out an assisted review project, our eDiscovery team will then guide the reviewers through the necessary steps to achieve the desired outcome.

The Complete Process

Control Set
The chosen review platform must be capable of measuring the accuracy of the assisted review project. This measurement entails using a control or truth set, which is a statistically significant, random sample taken from the dataset. We batch out these documents for manual review and simply code them as either ‘Relevant’ or ‘Not Relevant’. The results of this round serve as a marker for determining the F1 measure, a calculation used to monitor the stability of the project.
Pre-coded Seed Round
Training the system to code documents effectively can take time. In cases where a manual review has already carried out on a set of documents, we can use these documents as pre-coded seeds. By using the same designation field created for the assisted review project, we can use these documents as examples of ‘Relevant’ and ‘Not Relevant’ documents. Consequently, we can categorise more documents in the assisted review project because of this.
Training Rounds
We conduct training rounds to teach the system how to categorise documents, batch out a sample for manual review. This is ideally done by someone familiar with the case, and code them as 'Relevant' or 'Not Relevant'. Additionally, documents can be tagged as examples by checking a 'Use as Example' box. If they contain a substantial amount of text and are considered good examples. Alternatively, an extract of text can be copied and pasted into a 'Use Text Excerpt' text box.
Previous slide
Next slide

Quality Control Round

Therefore, upon completing training rounds, we execute a quality control round on the categorised documents to test the accuracy of the system’s grouping. We batch out a sample of the documents already categorised as a result of the training rounds for manual review. Then, the system can compare how many documents it categorised correctly and how many it ‘overturned.’ An ‘overturn’ occurs when the system categorises a document, for example, as ‘Relevant,’ and a manual review codes the document as ‘Not Relevant.’ Additionally, we can measure and analyse the number of ‘overturns’ to identify and correct issues within the TAR project.

Training Rounds & Quality Control

In summary, the aim of the training rounds and QC rounds is to categorise as many documents as possible, as accurately as possible, in line with the F1 score (a measure of a test’s accuracy), agreed at the start of the project. The end result produces a set of coded documents that facilitate timely production. These documents can further prioritise review by pushing the ‘Relevant’ documents to the review team first or batching out all the ‘Relevant’ documents for review. Therefore, whatever the reason for your assisted review project, predictive coding will play an increasingly crucial role in the future as data sizes grow and manual document review costs soar. TAR has quickly become an essential tool in litigations, prioritising documents for review while reducing time and overall costs.

Explore our complete eDiscovery solutions suite here. 

Contact Us

TechTalks

Navigate Tomorrow's Evidence Today.

Accreditations & partners

Sign Up to Our Mailing List | CYFOR Legal
CYFOR Legal | eDiscovery