Question answering for policy documents

Abstract

Privacy policies are legally enforceable documents notifying consumers of how a company intends to interact with the consumer’s data, but modern policy documents prove to be too long and complex for the average user to meaningfully read through. Natural language processing shows the potential to help users better understand the consequences of their consent by allowing users to ask questions they find relevant. We review some approaches to question answering for privacy policies and analyze the types of questions users ask and the characteristics of the answers they expect.