A computational cartographic study on the merge nature of topics in stages of French

Abstract

We present a study in model comparison and model selection to explore cartographic theoretical proposals and test their generalization ability. Specifically, we test whether an asymmetry arises between stages of French with respect to the merge nature of topicalized elements, as proposed in Wolfe (2022). We explore quantitative and simple computational methods to test the model’s predictions compared to a series of control group models. In particular, we evaluate predictions with respect to intervention locality effects. We explore seven treebanks, representing three stages of French (Old French, Middle French, and contemporary French). Our results partially confirm Wolfe’s (2022) proposal. Additionally, two key elements emerge: one concerning the type of locality involved in movement, and the other relating to a cartographic approach to the syntax of genres/registers.