Why is strategically including a Technology Assisted Review (TAR) expert in your review often the right move? The truth is that the choices available to lawyers in terms of TAR tools and their settings, combined with the naturally high variability in real-world documents’ structure and content, can sometimes make it difficult for an overtaxed review team to use TAR to its full advantage.
Nearly all TAR solutions, regardless of the underlying technology, have more levers to pull than are made available to the end user, and their functions are not always obvious. Most algorithms for clustering, machine learning, semantic parsing, and the like come with a profusion of input parameters, and choosing the best settings often requires an in-depth understanding not only of the data that’s been collected, but of the algorithms themselves.
From the computational linguist’s point of view, there are two ways to tackle the issue when packaging up a “raw” algorithm into a user-facing offering: present the review team with a dizzying array of choices to be made, or choose a sane default for each setting and keep it out of view. While either approach can easily speed a review while increasing recall and precision, neither unleashes the full power of a well-tuned TAR solution.
Understanding the quirks of the document population, and how they’re likely to interact with a particular algorithm and all its settings, is just as important. Appearances to the contrary, the TAR world is full of opportunities to leverage human intuition. An experienced, knowledgeable TAR expert brings both case knowledge and professional experience to bear when spotting text and metadata trends, which are then used to select an appropriate algorithm and a corresponding approach to the review.
Conduent offers TAR solutions to fit any review, including the ability to leverage a selection of algorithms and approaches, in-depth metadata analysis, and expert consulting services.
About the Author