Mapping the meaning of human dignity at the European Court of Human Rights: An unsupervised learning approach

This paper applies topic modeling techniques to trace the legal concept of human dignity in the jurisprudence of the European Court of Human Rights. Using unsupervised learning methods, we aim to detect recurring topics and themes in the case law on human dignity, offering a data-driven perspective to a long-standing theoretical debate. We implemented a specific preprocessing pipeline to prepare the dataset for analysis and employed several state-of-the-art topic modeling algorithms, including LDA, LSI, NMF, and BERTopic. To evaluate the results, we used a measure of topic quality, combining topic coherence and topic diversity. Our findings suggest that a coherent ‘substantive’ notion of dignity can indeed be inferred from the Court’s case law with some degree of consistency. This paper contributes both methodologically, by demonstrating the efficacy of different approaches to topic modeling in legal contexts, and substantively, by deepening the understanding of how the concept of human dignity is interpreted in human rights law.