Computational Investigations of Multiword Chunks in Language Learning

Area of Science:

• Psycholinguistics
• Computational Linguistics
• Developmental Linguistics

Background:

• Second-language learners often do not achieve native-like proficiency, particularly in morphology and syntax.
• Prior research focused on cognitive and neural factors to explain differences between first- and second-language acquisition.
• This study investigates the role of linguistic unit granularity in language learning outcomes.

Purpose of the Study:

• To test the hypothesis that children rely more on multiword units than adults learning a second language.
• To explore the granularity of linguistic units in first- and second-language acquisition.
• To computationally model language learning to compare the utility of chunk-based knowledge.

Main Methods:

• Utilized a computational model of language learning, the Chunk-Based Learner.
• Employed large-scale, corpus-based computational modeling.
• Compared the effectiveness of multiword units in accounting for speech in first-language children and adult second-language learners.

Main Results:

• Findings suggest multiword units play a role in second-language learning.
• Adult second-language learners may acquire less useful chunks compared to child first-language learners.
• Adult second-language learners may rely less on chunks and acquire them differently than children.

Conclusions:

• Linguistic unit granularity, specifically the reliance on multiword chunks, differs between child first-language acquisition and adult second-language acquisition.
• Computational modeling provides a valuable tool for investigating the nature of linguistic units in language learning.
• Differences in chunk utilization may contribute to the varying proficiency levels observed in first- and second-language learners.