The PCFP and the Zipfian frequency distribution: The Median Threshold Hypothesis and French conjugation
Résumé
The effects of frequency on memory in general have long been documented. Furthermore,
Ambridge et al. (2015) has demonstrated that token frequency has an effect on linguistic
behaviour at large. In morphology, researchers (e.g. Bybee, 1995; Corbett et al., 2001; Herce,
2016) have studied its links with inflectional irregularities. Recently, Blevins et al. (2017) have
outlined the implications of this Zipfian frequency distribution (Zipf, 1932) on the mod-
eling of inflectional behaviour. But, despite the recognition of this effect, most studies in
this domain ignore frequencies in their analyses (e.g. Stump and Finkel, 2013; Stump, 2015;
Bonami and Boyé, 2014; Bonami, 2014; Beniamine, 2018; Pellegrini, 2023). Yet other works,
(e.g. Boyé, 2016; Herce, 2022; Sims-Williams, 2022), confirmed the importance of this effect
on the predictability of inflected forms (a.k.a the Paradigm Cell Filling Problem, PCFP, Ack-
erman et al. 2009).
In this paper, we propose to test the impact of the token frequency on form predictions
based on a reference database of inflected forms, Flexique (Bonami et al., 2014) and a cor-
pus extracted lexicon with token frequencies, Lexique4Linguists (Schalchli, 2022) based on
OpenSubtitles2016 (Lison and Tiedemann, 2016).
Domaines
LinguistiqueOrigine | Fichiers produits par l'(les) auteur(s) |
---|