Legal Definition Annotation in EU Legislation Using Symbolic AI

The definition is an integral component of the legislation. It is essential to the adequacy and legitimacy of legislation. In legislation, legal definitions are used for clarity, consistency, and legitimate certainty, but also for creating new legal concepts (e.g., personal data) or crimes (e.g. stalking, mobbing). With the advancement in society with technological innovation, the significance of accurate and precise definitions in legislation is indeed more articulated. In order to avoid ambiguity and to ensure, as far as possible, a strict interpretation of Law, Legal Texts (LT) usually define the specific lexical terms used within their discourse by means of normative rules. Due to the continuous increase of LT and a large number of domain-specific rules, extracting these definitions from the LT would be costly and time-consuming if it's done humanely. Definition extraction is widely used in Legal domains to perform legal and compliance analysis. In this paper, we detect and annotate the legal definitions using Symbolic Artificial Intelligence (AI) based on Natural Language Processing (NLP) and fostering LegalXML annotation. The goal is to qualify a very valuable part of legislation for supporting further AI applications also in the judiciary domain. The detection and annotation of definitions are performed on the delimiting type of definitions. The dataset consists of EU Legislation in the span of time from 2010 to 2021 in Akoma Ntoso (AKN) file format. The resultant 15082 AKN files are annotated. (Akoma Ntoso OASIS LegalDocML XML Standard. http://docs.oasis-open.org/legaldocml/akn-core/v1.0/akn-core-v1.0-part1-vocabulary.html)