Dedre Gentner

Alice Gabrielle Twight Professor of Psychology & Education


Curriculum vitae



(847)467-1272


Department of Psychology

Northwestern University



Modeling structural priming in sentence production via analogical processes


Journal article


Jason Taylor, Scott E Friedman, Micah B. Goldwater, Kenneth D. Forbus, D. Gentner
Annual Meeting of the Cognitive Science Society, 2011

Semantic Scholar DBLP
Cite

Cite

APA   Click to copy
Taylor, J., Friedman, S. E., Goldwater, M. B., Forbus, K. D., & Gentner, D. (2011). Modeling structural priming in sentence production via analogical processes. Annual Meeting of the Cognitive Science Society.


Chicago/Turabian   Click to copy
Taylor, Jason, Scott E Friedman, Micah B. Goldwater, Kenneth D. Forbus, and D. Gentner. “Modeling Structural Priming in Sentence Production via Analogical Processes.” Annual Meeting of the Cognitive Science Society (2011).


MLA   Click to copy
Taylor, Jason, et al. “Modeling Structural Priming in Sentence Production via Analogical Processes.” Annual Meeting of the Cognitive Science Society, 2011.


BibTeX   Click to copy

@article{jason2011a,
  title = {Modeling structural priming in sentence production via analogical processes},
  year = {2011},
  journal = {Annual Meeting of the Cognitive Science Society},
  author = {Taylor, Jason and Friedman, Scott E and Goldwater, Micah B. and Forbus, Kenneth D. and Gentner, D.}
}

Abstract

Modeling structural priming in sentence production via analogical processes Jason L. M. Taylor ([email protected]) Scott E. Friedman ([email protected]) Kenneth Forbus ([email protected]) Qualitative Reasoning Group, Northwestern University, 2133 Sheridan Road Evanston, IL 60208 USA Micah Goldwater ([email protected]) Dedre Gentner ([email protected]) Department of Psychology, Northwestern University, 2029 Sheridan Road Evanston, IL 60208 USA Abstract Recently there has been a surge of interest in using structural priming to examine sentence production. We present an analogical model of sentence production that exhibits structural priming effects. It uses analogical generalization to acquire abstract language patterns from experience. To construct utterances, it uses analogical retrieval to find semantically similar utterances and generalizations, and constructs a new sentence by analogy to them. Using the stimulus generator of Chang et al (2006), we show that this model can exhibit structural priming effects similar to those observed in humans, but with orders of magnitude less prior experience than required by a previous simulation. Keywords: structural priming; sentence production; syntax acquisition; analogy. Introduction What mechanisms underlie sentence production? In particular, how do speakers choose among the multiple grammatical forms that are capable of expressing something they intend to convey? Recently, there has been a surge of interest in structural priming as a way to examine sentence production processes in adults and children (Bock, 1986; Bock & Griffin, 2000; Chang, Dell, & Bock, 2006; Kaschak & Borreggine, 2008; Savage et al., 2003). In structural priming, the structure of one sentence is repeated in the structure of a second sentence (Bock, 1986). Structural priming occurs without any intention to create syntactic parallelism. It does not require semantic or thematic overlap between the utterances, although the effects can be stronger when lexical items are repeated, and sentences are semantically similar (Branigan, Pickering, & Cleland, 2000; Goldwater et al., 2011; Hare & Goldberg, 1999; Pickering & Garrod, 2004; Snider, 2009). To illustrate, consider a scene of a man giving cake to his son. It could be described either by a double object dative construction (DO), as in 1, or by a prepositional dative construction (PP), as in 2. 1. The man gave his son some cake. 2. The man gave some cake to his son. If an experimenter describes this scene with the DO (the prime utterance), and then shows a picture of a girl telling her friend a story, structural priming would be shown by the increased likelihood that the participant's description of the scene (the target utterance) would use a DO as in 3 (rather than a PP as in 4). 3. The girl told her friend a story 4. The girl told a story to her friend. Structural priming is seen as evidence of abstract syntax because it can operate across semantically different utterances and across intervening sentences (Bock, 1986; Chang, Bock, & Goldberg, 2003; Thothathiri & Snedeker, 2008). Thus the development of structural priming in children has been used to mark the development of syntactic abstraction (e.g., Savage et al., 2003). Indeed, Chang et al. (2006) have suggested that the mechanisms underlying structural priming are the same mechanisms involved in learning grammar. Pickering and Garrod (2004) additionally suggest that structural priming is one mechanism by which conversational fluency between interlocutors is achieved. Two highly influential models, by Chang et al. (2006) and by Pickering & Garrod (2004), each account for many of the phenomena of structural priming. However, research by Goldwater et al. (2011) shows that some phenomena of structural priming can best be captured by using the mechanism underlying analogical reasoning—structure- mapping (Gentner, 1983; Gentner & Markman, 1997). We propose that structural priming can be modeled as a species of analogy. This proposal might seem surprising, given that analogy is often considered to be a conscious phenomenon while priming is clearly implicit. However, recent results show that structure-mapping from a prior analog can occur without attention or awareness (Day & Gentner, 2007). We describe an initial computational model, based on analogical processes of matching, generalization, and retrieval. To provide a solid basis for comparison, we use the experimental design and stimulus generator developed by Chang et al (2006). We begin by summarizing the psychological experiments and the Chang et al (2006) model. We then describe our analogy-based simulation, including its structure and operation. The results of three simulation experiments are presented.


Share



Follow this website


You need to create an Owlstown account to follow this website.


Sign up

Already an Owlstown member?

Log in