Words with Consistent Diachronic Usage Patterns are Learned Earlier: A Computational Analysis Using Temporally...

Area of Science:

• Computational linguistics
• Psycholinguistics
• Historical linguistics

Background:

• Language change is a complex phenomenon influenced by various factors.
• Understanding the relationship between language evolution and individual acquisition is crucial.
• Existing methods for quantifying language change often lack scalability or are not grounded in actual language use.

Purpose of the Study:

• To quantify language change in a data-driven and scalable manner using temporally aligned word embeddings.
• To investigate the relationship between measures of language change and age of acquisition (AoA).
• To explore different facets of language change by analyzing lexical representations and their relational dynamics.

Main Methods:

• Utilized a large diachronic corpus of English.
• Employed temporally aligned word embeddings to model lexical representations over time.
• Developed and applied measures of language change, considering both representational shifts and relational changes.
• Controlled for confounding variables such as word frequency, contextual diversity, and word length.

Main Results:

• Identified a unique and reliable relationship between language change metrics and AoA.
• Demonstrated that different measures of language change capture distinct aspects of linguistic evolution.
• Found that words with more coherent diachronic usage patterns are acquired earlier.
• Observed a stronger correlation between AoA and neighborhood-level measures of language change.

Conclusions:

• Language change is demonstrably linked to the age at which words are acquired.
• The findings support theories connecting ontogenetic (individual) and ethnogenetic (group) language development processes.
• This data-driven approach provides a scalable method for studying language evolution grounded in usage.