Coping with formulaicity in diachronic corpora

Abstract

This study explores ways in which formulaicity affects the diachronic linguistic analysis of historical text corpora. We show that the massive presence of formulaic language in an early medieval Latin charter corpus limits the diachronic conclusions that can be drawn from lexically determined corpus-linguistic measures. Focusing on diachronic productivity metrics, we suggest a way to mitigate the formula bias by means of adjustment coefficients calculated on the overall productivity rate of the entire vocabulary within the same diachronically successive subcorpora across which the productivity change of the linguistic constructions under examination is measured. We test our method on a set of Latin constructions which are known to be emergent, stable, or recessive, that is, shrinking over time, in the early Middle Ages. Our comparisons suggest that the diachronic profiles based on adjusted values, especially the adjusted neologism-based productivity , reflect the expected trends of the constructions better than unadjusted profiles.