Modeling the Acquisition of Words with Multiple Meanings
- Libby Barak (Princeton University)
- Sammy Floyd (Princeton University)
- Adele Goldberg (Princeton University)
Abstract
Learning vocabulary is essential to successful communication. Complicating this task is the underappreciated fact that most common words are associated with multiple senses (are polysemous) (e.g., baseball cap vs. cap of a bottle), while other words are homonymous, evoking meanings that are unrelated to one another (e.g., baseball bat vs. flying bat). Models of human word learning have thus far failed to represent this level of naturalistic complexity. We extend a feature-based computational model to allow for multiple meanings, while capturing the gradient distinction between polysemy and homonymy by using structured sets of features. Results confirm that the present model correlates better with human data on novel word learning tasks than the existing feature-based model.
Keywords: Language acquisition, Computational Modeling, Cross situational learning, Polysemy
How to Cite:
Barak, L., Floyd, S. & Goldberg, A., (2019) “Modeling the Acquisition of Words with Multiple Meanings”, Society for Computation in Linguistics 2(1), 216-225. doi: https://doi.org/10.7275/tr21-m273
Downloads:
Download PDF