Sentence-based attentional mechanisms in word learning: evidence from a computational model

Area of Science:

• Cognitive Science
• Developmental Psychology
• Computational Linguistics

Background:

• Word learning relies on cross-situational statistics and linguistic context.
• Previous research shows complex interactions between these two learning mechanisms.
• Adults' word learning is influenced by both statistical patterns and sentence cues.

Purpose of the Study:

• To model the interaction between cross-situational statistics and sentential context in word learning.
• To investigate how these two learning mechanisms influence each other.
• To examine the developmental trajectory of cue utilization in word acquisition.

Main Methods:

• Developed a probabilistic computational model of word learning.
• Extended a cross-situational model with a sentential cue-based attention mechanism.
• Simulated experiments from Koehne and Crocker (2010, 2011).

Main Results:

• The model replicates findings on the complex interaction between cross-situational and contextual word learning.
• Learning patterns emerge from probabilistic interactions, without explicit cue priority.
• The model allows examination of how cue roles change during learning.

Conclusions:

• A probabilistic computational approach can explain the interplay of different word learning mechanisms.
• Word learning is a dynamic process shaped by the probabilistic interaction of various cues.
• Computational models offer insights into the developmental aspects of language acquisition.