A Map of Constitutions: Exploring the Relation between Textual Similarities, Institutional Arrangements, and Legal Cultures

We develop innovative natural language processing methods to algorithmically match similar constitutional provisions on the basis of textual similarities, and employ them in conjunction with matching algorithms from computational social choice to arrive at the aggregate index of the textual similarity of constitutional documents. On the basis of this index we embed constitutional documents into a high-dimensional vector space, and thereafter use dimensionality reduction methods to map that space to a low-dimensional map suitable for exploratory analysis of factors affecting textual similarities of national and subnational constitutions. In particular, we will test two basic hypotheses: that textual similarities correlate with institutional arrangements and with legal cultures, while controlling for linguistic, historical, and economic similarities.