Empirical evidence for paradigmatic organization across morphology

Speaker:
Olivier Bonami (Université de Paris & CNRS)
Abstract:
There are two competing theoretical views on the nature of morphology, with deep implications both for descriptive and computational linguistics (Haspelmath & Sims, 2010; Blevins, 2016). On the one hand, morphology can be seen as grounded in primitives such as roots, affixes, and morphological processes; the task of the morphologist is then to make sense of the internal structure of “complex words”. The alternative views takes words to be primitives, and the task of the morphologist to study relations between these words – as Matthews (1991) put it, words then “are not wholes composed of simple parts, but are themselves the parts within a complex whole”.

In this talk I will first review empirical evidence collected over the last decade arguing in favor of the latter view, where paradigmatic relations between words are seen as central to morphology (Bonami & Beniamine, 2016; Bonami & Strnadová, 2019; Copot & Bonami, submitted). The evidence is based on behavioral and computational studies of both inflection and derivation, from the point of view of form predictability. I will then go on to present in more detail a recent study (Bonami & Guzmán Naranjo, in press) providing empirical evidence that paradigmatic relations play a substantial role for the predictability of meaning. Specifically, we rely on methods form distributional semantics to establish that the meaning of a complex word is sometimes easier to predict from a member of its morphological family other than its base.
Length:
01:05:19
Date:
27/02/2023
views: 279

Images:
Attachments: (video, slides, etc.)
60.0 MB
280 downloads
73.0 MB
161 downloads