Zero-Shot Cross-Situational Learning for Building Word-Referent Mappings
eScholarship (California Digital Library) · 18.06.2025
DOI: 10.48448/3mg0-dq79 ↗Kurz gefasst
Fachartikel in der Zeitschrift eScholarship (California Digital Library), erschienen am 18.06.2025. Verfasst von Cognitive Science Society 2025 und 2 weiteren Autor:innen.
Abstract (Originalsprache)
Statistical learning (SL) mechanisms drive learning across several domains, including language acquisition. Can SL mechanisms like cross-situational word learning, which rely on the accumulation of statistical evidence, also account for the rapid acceleration in children’s word learning? This study examined whether learners could track and integrate several statistical regularities concurrently and use this information to map words to objects that never co-occurred during training—a behavior known as zero-shot learning. Experiment 1 showed that learners leveraged their acquired knowledge of word and object categories to map words to objects that did not co-occur during training. Experiment 2 extended this finding by demonstrating that learners integrated their newly acquired statistics with prior knowledge to map novel words to perceptually similar, yet previously unseen, objects. These findings suggest that integrating several types of statistical relationships between words and objects can accelerate new learning, making SL an efficient mechanism for early word learning.
Schlagwörter
Zitieren
APA
2025, C. S. S., Knabe, M. L., & Yu, C. (2025). Zero-Shot Cross-Situational Learning for Building Word-Referent Mappings. eScholarship (California Digital Library). https://doi.org/10.48448/3mg0-dq79
Vancouver
2025 CSS, Knabe ML, Yu C. Zero-Shot Cross-Situational Learning for Building Word-Referent Mappings. eScholarship (California Digital Library). 2025. doi:10.48448/3mg0-dq79
BibTeX
@article{2025,
author = {2025, Cognitive Science Society and Knabe, Melina Lauryn and Yu, Chen},
title = {Zero-Shot Cross-Situational Learning for Building Word-Referent Mappings},
journal = {eScholarship (California Digital Library)},
year = {2025},
doi = {10.48448/3mg0-dq79}
}
EUVIMED