A Bootstrapping Model of Frequency and Context Effects in Word Learning.

Area of Science:

• Cognitive Science
• Developmental Psychology
• Computational Linguistics

Background:

• Cross-situational word learning involves mapping words to objects from ambiguous contexts.
• Successful learning relies on tracking word-referent co-occurrence frequencies.

Purpose of the Study:

• To investigate how varying frequencies of word-referent pairs impact learning.
• To explore if high-frequency pairs can facilitate the learning of low-frequency pairs.
• To examine the role of contextual diversity in word acquisition.

Main Methods:

• Utilized an associative model (Kachergis, Yu, & Shiffrin, 2012) to simulate learning.
• Manipulated the frequency of word-referent pairs during training.
• Varied the contextual diversity (number of other pairs a given pair appears with).

Main Results:

• High-frequency word-referent pairs do not always lead to better learning outcomes.
• The frequency of certain pairs can enhance the learning of other, less frequent pairs.
• An associative model successfully explains how frequency and contextual diversity interact to affect learning.

Conclusions:

• The interaction between familiarity and uncertainty biases in associative models explains learning patterns.
• The proposed associative model accounts for human learning data better than alternative models.
• Frequency and contextual diversity are crucial factors in naturalistic word learning environments.