Integrating a bottom–up and top–down methodology for building semantic resources for the multilingual legal domain

This article presents a methodology for multilingual legal knowledge acquisition and modelling. It encompasses two comlementary strategies. On the one hand, there is the top–down definition of the conceptual structure of the legal domain under consideration on the basis of expert jugdment. This structure is language–independent, modeled as an ontology, and can be aligned with other ontologies that capture similar or complementary knowledge, in order to provide a wider conceptual embedding. Another top–down approach is the exploitation of the explicit structure of legal texts, which enables the targeted identification of text spans that play an ontological role and their subsequent inclusion in the knowledge model. On the other hand, the linguistically motivated, text-based bottom–up population and incremental refinement of this conceptual structure using (semi-)automatic NLP techniques, maximizes the completeness and domain-specificity of the resulting knowledge. The proposed methodology is concerned with the relation between these two differently derived types of knowledge, and defines a framework for interfacing lexical and ontological knowledge, the result of which offers various perspectives on multilingual legal knowledge. Two case-studies combining bottom-up and top-down methodologies for knowledge modelling and learning are presented as illustrations of the methodology.