Density and Distinctiveness in Early Word Learning: Evidence From Neural Network Simulations.

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