Recovering from accidental syncretisms with word2vec analogies
Résumé
Systematic syncretism has received a lot of attention since Aronoff (1994) and the introduction of mor-
phomes. Morphologists have questioned its stability (Maiden, 1992), its orientation (Baerman et al., 2005), its
structure (Bonami and Boyé, 2003). With the emergence of the paradigm cell filling problem, PCFP, in (Ack-
erman et al., 2009) as formulated below in (1), systematic syncretism and the morphomes associated with it
have become part a central part of the discussion.
(1) What licenses reliable inferences about the inflected (and derived) surface forms of a lexical item?
With a morphomic paradigm (Boyé and Schalchli, 2019), systematic syncretisms are captured by equating
contrasting forms and their morphosyntactic distributions. In the absence of accidental syncretism, a form
defines a morphomic cell. For example, in English, the 8 different forms of BE define the cells of the morphomic
paradigm for all verbs. In this perspective, cells are not defined by a single set of morphosyntactic properties
being realized as a form but rather by a form associated with the contexts in which it appears, a distribution.
For example, in an ideal system, the distribution of the forms of BE would provide a set of key distributions
with which to classify the distribution of the forms of any other verb. But when accidental syncretism oc-
curs, this straightforward association between a verb form and one key distribution is obscured. Rather than
corresponding to a key, the syncretic form distribution results from the union of two or more key distributions.
In this paper, we propose to look at what we call the paradigm cell association problem (PCAP).
(2) Assuming that speakers have identified a morphomic paradigm shape based on contrasting forms, how
can they place forms in the morphomic paradigm?
We propose speakers can use the key forms they have already placed and compare their distribution with the
ones under scrutiny using all the combinations possible to identify accidental syncretism.
Domaines
LinguistiqueOrigine | Fichiers produits par l'(les) auteur(s) |
---|