Reverse Engineering Language Acquisition with Child-Centered Long-Form Recordings

Abstract

Language use in everyday life can be studied using lightweight, wearable recorders that collect long-form recordings—that is, audio (including speech) over whole days. The hardware and software underlying this technique are increasingly accessible and inexpensive, and these data are revolutionizing the language acquisition field. We first place this technique into the broader context of the current ways of studying both the input being received by children and children's own language production, laying out the main advantages and drawbacks of long-form recordings. We then go on to argue that a unique advantage of long-form recordings is that they can fuel realistic models of early language acquisition that use speech to represent children's input and/or to establish production benchmarks. To enable the field to make the most of this unique empirical and conceptual contribution, we outline what this reverse engineering approach from long-form recordings entails, why it is useful, and how to evaluate success. Expected final online publication date for the Annual Review of Linguistics, Volume 8 is January 2022. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.

A systematic review suggests marked differences in the prevalence of infant‐directed vocalization across groups of populations

“…Building on this idea, Lavechin et al. (2021) suggest that children's productions in long‐form recordings are benchmarked against those of adults in the same recordings. Ideally, such a move would help us avoid being trapped by inappropriate assumptions and unfair between‐population comparisons – although we recognize that this is never easy (Broesch et al., 2020).…”

“…Dupoux (2018), in particular, proposes a reverse-engineering approach, where computer models learn language from samples of children's everyday input as captured using long-form recordings. Building on this idea, Lavechin et al (2021) suggest that children's productions in long-form recordings are benchmarked against those of adults in the same recordings.…”

Introducing Meta‐analysis in the Evaluation of Computational Models of Infant Language Development

“…Time is also ripe for such efforts, as recent advances in machine learning enable the exploration of this path toward more comprehensive approaches attempting to learn at multiple levels with constraints and mechanisms plausible also for human learners (e.g., Nguyen et al, 2021;Alishahi et al, 2021). Importantly, the use of modern machine learning for modeling allows testing of these models on data of unprecedented scale and naturalness (e.g., out-of-lab long-form audio recordings or head-mounted camera data captured from infants living their daily lives; Lavechin, de Seyssel, Gautheron, Dupoux, & Cristia, 2022;Sullivan, Mei, Perfors, Wojcik, & Frank, 2021), allowing us to tackle an array of research questions related to language input and learning mechanisms that cannot be easily addressed with more simplistic data.…”

It Makes a Village: Child Care and Prosociality *

“…The number of child vocalizations was estimated from child-centered long-form recordings collected using a wearable device, an increasingly used technique in early language development research (Lavechin et al, 2021). 22 Each recorded child wore a t-shirt fitted with two small breast-pockets, into which a pair of USB voice recorders was inserted.…”