Automated Legislative Drafting : Generating

In this paper, we describe which roles deep structures of law play in drafting legislation. Deep structures contain a formal description of the intended normative eeects of a new regulation. We will discuss mechanisms that can be used to generate diierent paraphrases of regulations automatically. Since it is possible to test the paraphrases on legal knowledge based systems, we have provided two extra design steps in legislative drafting which can be supported by automated tools. Deep structures are straightforward descriptions of the normative eeects of regulations. Each deep structure distinguishes desired and undesired behaviour, and has no further internal structure such as paragraphs or exception structures. This paper describes methods to translate a deep structure to representations of diierent types of codes, i.e. paraphrases. Each representation of a code has a diierent structure, according to the choice we make during the translation regarding: a) the initial assumptions of the regulation, i.e. modelling from desired or undesired behaviour, b) the level of abstraction, c) the viewpoint of the law, i.e. the category of norm subjects, and d) the type of deontic modalities the regulation largely uses. All paraphrases have the same eeects as the deep structure, but with diierent features, and are suitable for diierent goals.