Describing Vocalizations in Young Children: A Big Data Approach Through Citizen Science Annotation

Abstract

Purpose Recording young children's vocalizations through wearables is a promising method to assess language development. However, accurately and rapidly annotating these files remains challenging. Online crowdsourcing with the collaboration of citizen scientists could be a feasible solution. In this article, we assess the extent to which citizen scientists' annotations align with those gathered in the lab for recordings collected from young children. Method Segments identified by Language ENvironment Analysis as produced by the key child were extracted from one daylong recording for each of 20 participants: 10 low-risk control children and 10 children diagnosed with Angelman syndrome, a neurogenetic syndrome characterized by severe language impairments. Speech samples were annotated by trained annotators in the laboratory as well as by citizen scientists on Zooniverse. All annotators assigned one of five labels to each sample: Canonical, Noncanonical, Crying, Laughing, and Junk. This allowed the derivation of two child-level vocalization metrics: the Linguistic Proportion and the Canonical Proportion. Results At the segment level, Zooniverse classifications had moderate precision and recall. More importantly, the Linguistic Proportion and the Canonical Proportion derived from Zooniverse annotations were highly correlated with those derived from laboratory annotations. Conclusions Annotations obtained through a citizen science platform can help us overcome challenges posed by the process of annotating daylong speech recordings. Particularly when used in composites or derived metrics, such annotations can be used to investigate early markers of language delays.

Auditory chaos classification in real-world environments

“…Additionally, we note that real-world models with accuracies in the range of 0.60 have made real contributions to empirical research in child development. For example, measures of overheard speech derived from LENA’s speech classifier which has an overall weighted accuracy of 67% ( 79 ) have predicted various measures of language development in young children. A review paper provides a summary of works that used LENA’s in-built algorithms to detect aspects of the speech environment and were found to significantly predict individual differences in child language development as well as gold-standard laboratory measures ( 80 ).…”

It Makes a Village: Child Care and Prosociality *

“…However, this also means that the audio recording is harder to process than an audio recording gathered in more manicured and stable conditions. Automatized algorithms attempt to classify child vocalizations into crying, laughing, canonical, and noncanonical, but precision is still a challenge (Schuller et al, 2019;Semenzin et al, 2021). We address the possibility that our vocalization counts confound actual speech and crying in several ways.…”

“…Among a sample of either children younger than children in our sample (4-18 months) or slightly older children (11-53 months) but who were diagnosed with Angelman syndrome (a genetic disorder causing speech delays and intellectual disability), related research has shown that 72.89% of automatically identified segments were speech-like, 5.23% crying, and 1.65% laughing(Semenzin et al, 2021).…”