Cleanlab for the Legal Industry

Cleanlab Studio was used by a customer to identify errors in relevance annotations for documents during the e-discovery phase of a legal proceeding. Cleanlab automatically identified a vast number of important documents that paralegals accidentally failed to annotate as relevant. Relevance-prediction models trained inside Cleanlab Studio were 15% more accurate than the customer’s own models. Similar benefits were observed for dealing with annotations for privileged content.