Mission Impossible?: Automated Norm Analysis of Legal Texts

1 Introduction Although many legal experts thought this would be impossible we are very close in creating an algorithm for automated norm analysis from legal texts. This algorithm makes use of invariant linguistic structures at the syntactical level that characterises specific normative expressions in natural language. Since the algorithm has not been realised and tested completely jet, we will limit ourselves in this article to explaining the invariances in the natural language representations in which norms in legal texts are expressed. As part of the POWER research program [1], the Dutch Tax and Customs Administration has created a method to formalise normative expressions in legal texts in UML/OCL models. These UML/OCL representations of the legal texts have showed to be quite suitable. To support knowledge analysts in creating these UML/OCL-models, an automated concept ex-tractor was created, which allows a computer to identify the different concepts that exist in a given legal text [2]. This automated concept extractor reduces the amount of work of the knowledge analysts and results in more uniform models as well. The research described here is aimed at further automating the translation of a legal text to a model. Automated generation of models would not only lead to a reduction in the amount of work needed, it would also increase inter-analyst independency. Normally, models created by different analysts could differ in various small details. Removing those difference would lead to more uniform models, which can more easily be understood, and are also easier to process when creating applications based on these models. This article discusses the first results of this research into automated analysis of legal texts.