Technology Assisted Reviews

Six Practical Tips To Maximize The Use Of A Technology Assisted Review

The use of technology in assisted reviews is forthcoming. Its predecessor, machine learning strategies, is already around and being used in various industrialized applications such as speech recognition and face categorization. Technology assisted reviews are used to reduce costs, speed up reviews, and increase accuracy and efficiency. The review team takes a representative sample from the actual document set to program the computer. Since electronically-stored information is seeing an explosive growth, it makes logical sense to take advantage of a technology which will not only magnify the efforts of a single reviewer but also reduce the cost and time associated with an assisted review.

The following six suggestions will help the reader maximize the use of a technology assisted review:

1. The system should be trained viz subject matter experts

From the beginning to the end, the same set of reviewers should be involved in training the system. This will maintain consistency. A review team that is not familiar with the subject matter at hand, should not be used.

2. A manager who is trained with Technology Assisted Reviews should head the team

Large firms and corporations that are dealing with large or complicated cases, tend to outsource review teams. An outsourced management is more likely to produce accurate and expedient results. Assisted reviews are used for many purposes such as case assessments, bulk classification, and prioritizing document reviews. A successful technology assisted review results when the review team and review manager work as group.

3. Learn the Technology Jargon

There are some words which will frequently appear in a technology assisted review. These common terms include richness, recall, f-measure, and precision. When everyone on the team is familiar with the same vocabulary, it ensures accuracy and a speedy review.

4. Documents with primary issues should be ranked first

While reviewing documents, responsive documents should be subdivided according to primary concerns.  The review team should not focus on more than two primary issues at a time.

5. Implement blind random sampling to ensure accuracy and avoid bias

An in-built probability algorithm, acts as the quality control in a technology assisted review. If a reviewer is aware of which documents are marked as non-responsive or responsive, then it creates reviewer bias. Using a random sample from your actual document set will ensure that the results represent the population at large.

6. Second-pass linear review should be used

Generally, a second pass review is suggested for documents which are privileged or non-responsive.

The electronic storage of information is becoming a widely acceptable trend. Compared to other machine learning methods, technology assisted reviews are gaining more favor. A technology assisted review is the best choice because it increases efficiency, reduces cost, and saves time.