Abstract
A method, a system, and a computer program product for identifying sensitive data. A plurality of machine learning models are applied to an electronic document to generate a plurality of predictions. Each prediction identifies a location in a plurality of locations of sensitive data in the document. The plurality of locations of sensitive data in the electronic document are compared. Based on the comparison, one or more overlapping and non-overlapping locations of sensitive data in the electronic document are determined. At least one location in the plurality of locations corresponding to at least one of: the overlapping locations, the non-overlapping locations, and any combinations thereof is selected. A final prediction corresponding to each selected location of the sensitive data in the electronic document is generated.