Automated Compositional Change Detection in Saxo Grammaticus' Gesta Danorum

Saxo Grammaticus’ medieval source Gesta Danorum (“Deeds of the Danes”) represents the beginning of the modern historical research in Denmark. The bipartite composition of Gesta Danorum has however been subject to much academic debate. In particular the nature and location of a transition between early Pre-Christian and late Christian content have given rise to two competing accounts. In this paper, we argue that the debate can be represented as a problem of intratextual dynamics and we combine models from Information Retrieval and Natural Language Processing with techniques for time series analysis in order to reevaluate the debate. Results indicate that the transition is gradual, starting in book eight and ending in book ten, but that a point-like interpretation is possible in book nine. We argue that the approach exemplifies scalable“automated close reading”, which has multiple applications in text-based historical research. Keywords— Change Detection, Cultural Heritage, Medieval Literature, Text Analysis, Topic Modeling