Computer modelling of innovations relative to Latin in contemporary Romance dialects

This study relies on a corpus illustrating several dozen Romance dialects from France, Belgium, Switzerland, Italy, Spain and Portugal, for which 145 innovations relative to Latin have been encoded in the form of 1 (presence) or 0 (absence). Based on contemporary recordings (translations of Aesop’s 100-word fable “The North Wind and the Sun” and another 100-word list), following the principles of dialecto­metry, the Comparative method and especially historical glottometry, we propose computational tools to address the relationships and classifications amongst these Romance varieties. Results of data-mining techniques confirm the robustness of a North/​South divide — with the Oïl area being, by far, the most innovative — and, secondarily, an opposition between the South-West (mainly Ibero-Romance) and the South-East (mainly Italo-Romance, more conservative). Among the most important/​discriminant features are the palatalisation of Latin ca, which characterises the majority of northern Gallo-Romance dialects, and the simplification of geminates north of the La Spezia-Rimini line. Most innovations relate to phonetic/​phono­logical traits. However, we also consider morphosyntactic and lexical features, such as non-null subject in northern Gallo-Romance varieties, and the substitution of cum ‘with’ by apud > amb in Occitano-Romance varieties. By retaining only morpho­syntactic innovations, we still find a North vs. South-East- vs. South-West tripartition.