TY - JOUR

T1 - Density and distinctiveness in early word learning

T2 - Evidence from neural network simulations

AU - Jones, Sam

AU - Brandt, Silke

PY - 2020/1/9

Y1 - 2020/1/9

N2 - High phonological neighborhood density has been associated with both advantages and disadvantages in early word learning. High density may support the formation and fine-tuning of new word sound memories; a process termed lexical configuration (e.g. Storkel, 2004). However, new high-density words are also more likely to be misunderstood as instances of known words, and may therefore fail to trigger the learning process (e.g. Swingley & Aslin, 2007). To examine these apparently contradictory effects, we trained an autoencoder neural network on 587,954 word tokens (5497 types; including mono- and multi-syllabic words of all grammatical classes) spoken by 279 caregivers to English-speaking children aged 18 to 24 months. We then simulated a communicative development inventory administration and compared network performance to that of 2292 children aged 18 to 24 months. We argue that autoencoder performance illustrates concurrent density advantages and disadvantages, in contrast to prior behavioural and computational literature treating such effects independently. Low network error rates signal a configuration advantage for high-density words, while high network error rates signal a triggering advantage for low-density words. This interpretation is consistent with the application of autoencoders in academic research and industry, for simultaneous feature extraction (i.e. configuration) and anomaly detection (i.e. triggering). Autoencoder simulation therefore illustrates how apparently contradictory density and distinctiveness effects can emerge from a common learning mechanism.

AB - High phonological neighborhood density has been associated with both advantages and disadvantages in early word learning. High density may support the formation and fine-tuning of new word sound memories; a process termed lexical configuration (e.g. Storkel, 2004). However, new high-density words are also more likely to be misunderstood as instances of known words, and may therefore fail to trigger the learning process (e.g. Swingley & Aslin, 2007). To examine these apparently contradictory effects, we trained an autoencoder neural network on 587,954 word tokens (5497 types; including mono- and multi-syllabic words of all grammatical classes) spoken by 279 caregivers to English-speaking children aged 18 to 24 months. We then simulated a communicative development inventory administration and compared network performance to that of 2292 children aged 18 to 24 months. We argue that autoencoder performance illustrates concurrent density advantages and disadvantages, in contrast to prior behavioural and computational literature treating such effects independently. Low network error rates signal a configuration advantage for high-density words, while high network error rates signal a triggering advantage for low-density words. This interpretation is consistent with the application of autoencoders in academic research and industry, for simultaneous feature extraction (i.e. configuration) and anomaly detection (i.e. triggering). Autoencoder simulation therefore illustrates how apparently contradictory density and distinctiveness effects can emerge from a common learning mechanism.

U2 - 10.1111/cogs.12812

DO - 10.1111/cogs.12812

M3 - Journal article

VL - 44

JO - Cognitive Science

JF - Cognitive Science

SN - 0364-0213

IS - 1

M1 - e12812

ER -