Unlocking NLP techniques for non-quantitative researchers: regulatory aspects

Out of the major AI disciplines, Natural Language Processing (NLP) is possibly the most directly relevant to legal practice. Aiming to enable computers to process human language, the ability of NLP-based AI tools to take over legal tasks such as drawing up contracts or answering simple legal questions is what normally catches the eye. As such, the development of the burgeoning field of ‘NLP & Law’ tends to happen through academia-industry collaborations. This paper deals with two distinct, often overlooked, aspects of applying NLP techniques to law: 1. The use of advanced NLP techniques in academic research 2. The application of NLP techniques in a regulatory context Concretely it builds on experiences with WetSuite, a developing Dutch digital infrastructure aimed at unlocking NLP techniques for researchers in law and social science without programming skills. WetSuite covers a wider range of legal documents that commercial applications tend to do, specifically certain policy and soft law documents, regulatory decisions and complaints. This ‘unlocking exercise’ has the potential to generate a lot of new information that is relevant to the disciplines of law & regulation, such as with regard to consistency and bias in decision-making, changes in moral compass over time, intention vs. interpretation across branches of government, linguistic aspects of ‘legalese’, detecting problematic provisions in the law, compliance, and personalized access to administrative justice. However, it also comes with distinct legal, ethical and practical challenges. The paper identifies these and, where possible, proposes strategies for overcoming them.