NSP-SCD: A corpus construction protocol for child-directed print in understudied languages

Area of Science:

• Psycholinguistics
• Corpus Linguistics
• Child Language Acquisition

Background:

• Systematic psycholinguistic research requires child-directed print corpora, which are scarce for understudied languages.
• Researchers often rely on manual tagging due to the lack of automated parsers for these languages.
• Small corpora are a potential solution, but their robustness is not well-understood.

Purpose of the Study:

• To evaluate the effectiveness of a non-sequential sampling protocol for small corpus development (NSP-SCD).
• To assess the lexical and orthographic properties of a small corpus created using NSP-SCD.
• To compare the efficiency of NSP-SCD with traditional sampling methods for small corpus construction.

Main Methods:

• A small corpus (17,584 words) was created using NSP-SCD from a larger corpus (150,595 words) of children's books (ages 3-10).
• Cross-corpora and within-corpus analyses were performed to compare the small and large corpora.
• Lexical diversity, orthographic representation, word length, and psycholinguistic complexity were analyzed.
• Robustness was checked by comparing NSP-SCD with a same-sized corpus from a limited number of books.

Main Results:

• The small corpus approximated the lexical and orthographic diversity of the larger corpus.
• The small corpus was equivalent to the larger corpus in orthographic representation and word length.
• Psycholinguistic complexity varied by book level and part of speech.
• NSP-SCD was more efficient for lexical diversity than a same-sized corpus from few books.

Conclusions:

• Non-sequential sampling protocol for small corpus development (NSP-SCD) is a promising method for creating representative small corpora.
• Stratified sampling from books by level enhances corpus statistics, making them approximate larger corpora.
• This protocol can advance child language acquisition research in understudied languages by enabling more robust corpus creation.