Quantitative measures of child-directed speech across language communities.

BUNCE, JOHN; STEIN, A.; BERGELSON, E.; SODERSTROM, M.; ROSEMBERG, C.R.; CASILLAS, M.

Pensilvania

Conferencia; Conference: IASCL 2021 Biennial Meeting.; 2021

International Association for the Study of Child Language (IASCL)

Cross-cultural work lacks consistent methodologies, the existing automated tools (i.e., LENA) have a number of issues. ACLEW Goals●An annotation system for longform audio recordings?● Uses ELAN (free software) https://tla.mpi.nl/tools/tla-tools/elan/● Designed with non-experts in mind● Freely available tutorials in English, Spanish, and French, for Windows and for Macosf.io/b2jep/wiki/home/● All annotators certified with Gold standard test to ensure quality coding benchmarks● Every minute of annotation can be re-used for comparative analyses or tool developmentAclew Annotation SystemMeasures & Analytic Approach● Measures included:○ Quantity of speech (minutes per hour)○ Language Family (e.g., Arg. Spanish, Yeli Dnye)○ Addressee (Target Child-directed Speech (TCDS)/ Adult-directed speech (ADS))○ Speaker Type (Woman/Man/Child)○ Number of speakers present○ Child?s age (months)● Analyses used Zero Inflated Negative Binomial Mixed-effects Regression○ data constrained values >= 0Key Findings● The majority of child-directed speech in children?s environments comes from adult females.● The relative rates of adult-, target child-, and child-directed speech across communities arehighly variable.● Examine similarities and differences in child language environments within and acrosscultures, as well as influences on language ?outcomes?● Bring together individual corpora with a unified framework● Quantifiy language exposure across communities speaking North American English, UKEnglish, Argentinian Spanish, Yélî Dnye, and Tseltal Mayan● Provide gold standard data for the development of automated language processing tools.