To assess the current state of the field, we use four criteria to guide our discussion of prior work (see Table 1), and to make the comparisons more informative, we focus on the domain of early word learning rather than attempt to do a comprehensive review of connectionist models of language development. We will return to this point in the discussion and touch on other developmental modeling approaches. Although connectionism is not necessarily the only way to model these aspects of development (e.g., see Yu et al., 2005 Kemp et al., 2007 Xu and Tenenbaum, 2007 Frank et al., 2009), current research suggests that this is an especially promising approach. Connectionist models have the ability to capture processes of change over time as well as to capture multiple timescales of learning, all of which, as we argue in this review, makes them a good candidate model for development. The current review includes only connectionist models. This is because learning in connectionist models is incremental and representations are often under-determined in the beginning and learned as a way to solve a particular task. In general, connectionist models are well suited to model the time-course and emergent properties of processes. In the domain of language development, connectionist models have been used to help explain behavioral data, to test mechanistic accounts of language learning, and to inform big theoretical debates (e.g., Smith et al., 2010 Elman, 2011 Seidenberg and Plaut, in press). Throughout the paper, we will explore how computational models of word learning add insight to what is known about this developmental process as well as guide further discoveries.Ĭonnectionist models have made significant contributions to our understanding of various phenomena observed in young children (see Munakata et al., 2008 for a review). Our approach to modeling word learning has captured intriguing patterns of behavior, produced novel predictions, and has promise for exciting future applications. We then turn to our own work modeling developmental trajectories in typically developing children and late talkers. We argue that, however, most of these models do not fully take advantage of the strengths of connectionist models in capturing the temporality of development. We illustrate these characteristics by reviewing several prominent connectionist models of word learning. In this review we focus on the domain of semantic development, specifically early word learning, and highlight the characteristics of the connectionist approach that make it well-suited for modeling developmental processes. Nowhere is this feature more critical than in the modeling of developmental processes, which by definition occur in time. We argue that a modeling approach that truly captures change over time has the potential to inform theory, guide research, and lead to innovations in early language intervention.Īt the core of connectionist models is the idea of modeling change over time. We also discuss the potential of these kinds of models to capture children’s language development at the individual level. In this review paper, we discuss several prominent connectionist models of early word learning, focusing on semantic development, as well as our recent work modeling the emergence of word learning biases in different populations. However, fewer models of language development have truly captured the process of developmental change over time. Several models in the literature have made good contact with developmental data, effectively captured behavioral tasks, and accurately represented linguistic input available to young children. 2Department of Electrical, Computer, and Energy Engineering, University of Colorado Boulder, Boulder, CO, USAĬonnectionist models that capture developmental change over time have much to offer in the field of language development research.1Department of Psychology and Neuroscience, University of Colorado Boulder, Boulder, CO, USA.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |