Partial productivity of linguistic constructions: Dynamic categorization and statistical preemption

Abstract

Grammatical constructions are typically partially but not fully productive, which leads to a conundrum for the learner. When can a construction be extended for use with new words and when can it not? The solution suggested here relies on two complementary processes. The first is dynamic categorization: as learners record the statistics of their language, they implicitly categorize the input on the basis of form and function. On the basis of this categorization process, general semantic and phonological constraints on productivity emerge, and productivity is to a large extent determined by the degree to which the category is well attested by similar exemplars. Occasionally, a semantically sensical and phonologically well-formed instance of a well-attested construction is simply not fully acceptable. It is suggested that a process of statistical preemption is at work in these cases: learners avoid using a construction if an alternative formulation has been systematically witnessed instead. The mechanism proposed for statistical preemption is competition-driven learning: when two competitors are activated but one reliably wins, the loser becomes less accessible over time. In this way, the paradox of partial productivity can be resolved.

Shared neural and cognitive mechanisms in action and language: The multiscale information transfer framework

“…A schema can be thought of as a memory unit which generalizes across similarities, but abstracts away from differences between sequences of analogous structure. Schemas allow for some degree of flexibility regarding the concrete execution of sequences, thereby allowing for transfer and adaptation to novel contexts (Braun, Mehring, & Wolpert, 2010;Goldberg, 2016). Thus, experience in inline skating will facilitate the learning of iceskating, and experience with different sentences conveying a resultative meaning (e.g., He painted his house pink) will facilitate the generation of previously unheard, but structurally and functionally analogous sentences (e.g., Sally sneezed the napkin off the table; Goldberg, 1995, p. 6).…”

Wie viele Konstruktionen stecken in einem Wortbildungsmuster?

“…Die Frequenzverläufe in Abbildung 1 liefern einige Anhaltspunkte dazu, wel che konkreten Partizipien sich zu produktiven Subschemata herausbilden, aller dings erlauben sie keine Rückschlüsse auf Subschemata, an denen gleichzeitig mehrere, semantisch verwandte Partizipien beteiligt sind. Mit Goldberg (2016) kann die Frage gestellt werden, wie sich die Partizipien aus dem Datensatz im semantischen Raum verteilen und ob die englischen Partizipialkomposita einen Fall flächendeckender Streuung oder eher spärlicher Streuung darstellen. Um sich dieser Frage zu nähern, kann das Instrument einer semantischen Vektorraum analyse (Turney/Pantel 2010; Levshina 2015) benutzt werden, die es erlaubt, die im Datensatz vertretenen Partizipien im Hinblick auf ihre semantische Ähnlich keit hin zu vergleichen.…”

“…Es wäre wünschenswert, die Unterteilung eines Netzwerks in verschie dene Abstraktionsebenen und Subschemata auf eine solidere Basis zu stellen, die intersubjektiv reproduzierbare Ergebnisse liefert. Gries (2011) liefert dazu bereits einige Denkanstöße, die sich auf aktuelle Ansätze wie Goldberg (2016), Perek (2016) und den vorliegenden Beitrag anwenden lassen. Von einer weiteren kon struktionsgrammatischen Auseinandersetzung mit dem Begriff der Produktivität sind also interessante Ergebnisse zu erwarten.…”

Frequency vs. Association for Constraint Selection in Usage-Based Construction Grammar

“…This operationalizes the idea within usage-based theories of grammar that any representation can be stored in memory but that not all representations are worth storing (c.f., Jackendoff, 2002;O'Donnell, et al, 2011). From a different perspective, some constructions prevent the learning of other constructions (Goldberg, 2011;Goldberg, 2016;Perek & Goldberg, 2017). This paper first considers how constructions and slot-constraints can be represented computationally using a data-driven pipeline (Sections 2 & 3).…”

“…The frequency-based model represents what Goldberg calls conservatism via entrenchment, the idea that learners are more willing to overgeneralize infrequent forms (Goldberg, 2016). In other words, the problem with a frequency-based model is that it does not allow for creative (and thus infrequent) uses of common forms.…”