Diachronic syntax based on constituency and dependency annotated corpora

Abstract

This contribution presents two syntactically annotated corpora of Old French, Modéliser le changement: les voies du français (MCVF) and the Syntactic Reference Corpus of Medieval French (SRCMF). The focus is on how the underlying syntactic theory (constituency vs. dependency) influences the grammar model and how this choice is reflected in the syntactic annotations of the corpora. The comparison relates to the most relevant general properties of the corpora as well as to two phenomena, null subjects and cleft constructions. Null subjects highlight possible conflicts between syntactic annotation models and syntactic theory, and the information-structural properties of cleft constructions pose a particular problem for the interpretation and annotation of historical corpora. Both phenomena are major instances of diachronic variation in French. The study is relevant for corpus users working on diachronic syntax, as well for corpus builders wishing to design a grammar model for annotation.