This work presents a cognitive model of argument structure construction acquisition and application based on analogy. The claims of this model are that (1) constructions, pairings of form and meaning, are a productive unit of linguistic analysis that account for a broader range of phenomena than traditional approaches; (2) human acquisition of abstract argument structure constructions can be modeled as analogical generalization over individual examples; and (3) semantic interpretation can be modeled as the analogical integration of argument structure constructions and their constituents. In support of these claims, the primary contribution of this work is a computational model of argument structure acquisition and application that is built on top of the Structure Mapping Engine, a pre-existing computational model of analogy. ', '\tThe model was evaluated via three cognitive experiments. First, the model was used to simulate semantic inferences from Kaschak & Glenberg’s (2000) study of denominal verb interpretation which found that participants are influenced by argument structure when interpreting denominal verbs. The same model was also trained on child directed speech, with different parameters used to model conservative and liberal learners. The resulting constructions were analyzed for correctness and for an item-specific bias found in early language learning. Finally, the model was incorporated into the Analogical Theory of Mind model to simulate Hale & Tager-Flusberg’s (2003) study on linguistic bootstrapping effects in theory of mind acquisition. Taken together, these experiments provide evidence for an analogical approach to argument structure as a valid cognitive model.', '\tFurthermore, it is also claimed (4) that the model has practical applications for natural language tasks such as question answering. To evaluate this claim, the model was extended as a domain-general semantic parser. The resulting system was used to achieve high performance on seven of Facebook’s bAbI question answering tasks (Weston et al., 2015).

Last modified

• 10/14/2019

Creator

• Clifton James McFate

DOI

• https://doi.org/10.21985/n2-0a8j-p657

Subject

• Computer Science

Keyword

• Construction Grammar
• Analogy
• Cognitive Modeling

Date created

• 2018-01-01

Resource type

• Dissertation

Rights statement

• In Copyright