Mais et associés

Abstract

This paper focuses on the distribution of discourse marker sequences immediately on the left or right of mais (≈ but) in French, such as oui non mais (≈ yes no but), mais quand même (≈ but still) or mais bon (≈ but well). The goal is to determine what sequences are the most frequent and how they fit (or not) with the meaning of mais. Exploiting five spoken French corpora, we use two association measures to extract the most plausible candidate patterns. Following the literature on association measures, we explore the two complementary dimensions of frequency (MI3 measure) and predictability (DeltaP measure). This procedure reveals that i) there are indeed discourse marker patterns around bon, ii) most clusters or associates smoothly combine with the basic ‘argumentative’ value of mais and iii) within this semantically coherent set, comparing and crossing the results of the two measures on the left and the right of mais helps to identify subsets of discourse markers with specific discourse functions.