By Hema Begum
Legal firms and in-house counsel often face large data sets that must be reviewed and can tie up a significant amount of resource. We’re working with clients to drastically increase efficiency of the review process and cut costs by harnessing the power of AI.
At the heart of the eDiscovery offering is the advanced Active Learning and Propagation functions of our Relativity software. These technologies are subsets of machine learning and have transformed data review processes:
Active Learning
Active Learning is a review tool that uses machine learning to help organise your data and predict which documents are most likely to be relevant to reviewers. It works by learning from reviewers' coding decisions - for example where documents are marked "Relevant" and "Not Relevant", to prioritise the most appropriate ones to highlight for review. As reviewers code, the model gets better at selecting those that it needs to serve up to the reviewers. The aim is to review the most relevant documents first and cease the review when rates of relevance have dropped off and potentially before all documents have been reviewed.
Propagation
The propagation function can help enhance your document review workflow by saving time spent on coding documents. For example, you can tag a document as "Relevant" or "Not Relevant" and have that tag attributed to the document’s family members or duplicate document. When you code a document in a field, the propagation function automatically codes related documents with the same value which rapidly speeds up review processes.
What does this mean for lawyers?
Our technology, which is guided by experts in machine learning, can rapidly speed up reviewing processes and cut legal costs. This leaves you to free up and focus resources on the most value-adding tasks for your clients.
We’re now using AI/machine learning on most of our eDiscovery matters. With the launch of the Relativity aiR, the Generative AI product, which we also now offer, it’s going to be interesting to see how much more cost can be saved (particularly when weighed against its additional cost as a product). More on that in due course.
