Extracting Word Frequencies From Child Language Corpora: A New Tool for Developmental Psycholinguistic Research.

Area of Science:

• Linguistics
• Developmental Psychology
• Computational Linguistics

Background:

• Child language acquisition research relies on understanding linguistic input.
• Existing frequency databases often use adult-generated text, potentially misrepresenting child-directed speech.
• Analyzing word and phrase frequencies in child-directed speech is crucial for accurate developmental interpretations.

Purpose of the Study:

• Introduce childeswordfreq, an R package for extracting word and phrase frequency data from the Child Language Data Exchange System (CHILDES) database.
• Enable researchers to efficiently analyze linguistic input across languages and speaker roles within CHILDES.
• Facilitate stimulus selection for experimental studies and analysis of child language development patterns.

Main Methods:

• Developed childeswordfreq as an R-based tool for CHILDES data extraction.
• Conducted a case study using items from the Expressive One-Word Picture Vocabulary Test-Fourth Edition (EOWPVT-4).
• Compared EOWPVT-4 item frequencies in CHILDES with adult-based Subtitle-based Word Frequency Database for American English (SUBTLEX-US) estimates.

Main Results:

• Demonstrated patterned divergences between CHILDES and SUBTLEX-US frequencies, linked to developmental progression.
• Found earlier EOWPVT-4 items were overrepresented in child-directed speech (CHILDES).
• Observed later EOWPVT-4 items were more prominent in an adult corpus (SUBTLEX-US).

Conclusions:

• Childeswordfreq provides crucial access to child-directed speech frequency data for research.
• Adult-based frequency norms may not accurately reflect linguistic input relevant for child language acquisition.
• Utilizing child-based input patterns strengthens developmental interpretations and supports ecologically valid research.