A Computational Model of Early Argument Structure Acquisition

Abstract

How children go about learning the general regularities that govern language, as well as keeping track of the exceptions to them, remains one of the challenging open questions in the cognitive science of language. Computational modeling is an important methodology in research aimed at addressing this issue. We must determine appropriate learning mechanisms that can grasp generalizations from examples of specific usages, and that exhibit patterns of behavior over the course of learning similar to those in children. Early learning of verb argument structure is an area of language acquisition that provides an interesting testbed for such approaches due to the complexity of verb usages. A range of linguistic factors interact in determining the felicitous use of a verb in various constructionsassociations between syntactic forms and properties of meaning that form the basis for a number of linguistic and psycholinguistic theories of language. This article presents a computational model for the representation, acquisition, and use of verbs and constructions. The Bayesian framework is founded on a novel view of constructions as a probabilistic association between syntactic and semantic features. The computational experiments reported here demonstrate the feasibility of learning general constructions, and their exceptions, from individual usages of verbs. The behavior of the model over the timecourse of acquisition mimics, in relevant aspects, the stages of learning exhibited by children. Therefore, this proposal sheds light on the possible mechanisms at work in forming linguistic generalizations and maintaining knowledge of exceptions.

Do as I say, not as I do: A lexical distributional account of English locative verb class acquisition

“…In contrast to existing models of verb class acquisition (Alishahi & Stevenson, 2008;Niyogi, 2002;Perfors, Tenenbaum & Wonnacott, 2010) which assumed that children use adult-like semantics and syntactic features to acquire verb classes, the present work suggested that children can assign some verbs to classes based on word co-occurrence information. However, because language acquisition in the present model included a message with various kinds of meaning, it is important to characterise how our account differs from these accounts.…”

“…Various mechanistic accounts of how verb classes might be learned from semantic and syntactic information have been proposed (Alishahi & Stevenson, 2008;Niyogi, 2002;Perfors, Tenenbaum & Wonnacott, 2010). Learning in these models was quick, because their cues for verb bias were adult-like and optimal for solving this problem.…”

“…Instead, it specified hyper-parameters that constrained the types of hypotheses that were entertained and used adultlike syntactic structures and verb class semantic features to set these parameters. Alishahi and Stevenson's (2008) Bayesian model of verb argument structure acquisition extended these assumptions by enriching their syntactic structures with a large number of fine-grained semantic features (e.g., change of state) for learning. Because there is no automated way to label corpora for syntactic structure or fine-grained semantics, these models all took corpora that had been manually labelled with structural or semantic features by adult linguists.…”

A Usage-Based Model of Early Grammatical Development

“…Starting from initially small and concrete representations, it learns incrementally long representations (syntagmatic growth) as well as more abstract ones (paradigmatic growth). Several models address either paradigmatic (Alishahi and Stevenson, 2008;Chang, 2008;Bannard et al, 2009) or syntagmatic (Freudenthal et al, 2010) growth. This model aims to explain both, thereby contributing to the understanding of how different learning mechanisms interact.…”

A multimodal corpus for the evaluation of computational models for (grounded) language acquisition

“…those of lower complexity, as already solved by the learner. For instance, models addressing the acquisition of grammatical constructions and their meaning (Kwiatkowski et al, 2012;Alishahi and Stevenson, 2008;Gaspers and Cimiano, in press;Chang and Maia, 2001) typically learn from symbolic input. In particular, assuming that the child is already able to segment a speech signal into a stream of words and to extract structured representations from the visual context, such models typically explore learning from sequences of words and symbolic descriptions of the non-linguistic context.…”