Prespeech motor learning in a neural network using reinforcement

Abstract

Vocal motor development in infancy provides a crucial foundation for language development. Some significant early accomplishments include learning to control the process of phonation (the production of sound at the larynx) and learning to produce the sounds of one’s language. Previous work has shown that social reinforcement shapes the kinds of vocalizations infants produce. We present a neural network model that provides an account of how vocal learning may be guided by reinforcement. The model consists of a self-organizing map that outputs to muscles of a realistic vocalization synthesizer. Vocalizations are spontaneously produced by the network. If a vocalization meets certain acoustic criteria, it is reinforced, and the weights are updated to make similar muscle activations increasingly likely to recur. We ran simulations of the model under various reinforcement criteria and tested the types of vocalizations it produced after learning in the differ-ent conditions. When reinforcement was contingent on the production of phonated (i.e. voiced) sounds, the network’s post learning productions were almost always phonated, whereas when reinforcement was not contingent on phonation, the network’s post-learning productions were almost always not phonated. When reinforcement was contingent on both phonation and proximity to English vowels as opposed to Korean vowels, the model’s post-learning productions were more likely to resemble the English vowels and vice versa.

An iterative probabilistic model of speech-related vocalization rate growth due to child-caregiver interaction

“…It might therefore serve as a guide for future work on more detailed models in order to include realistic social response rates and learning rates while at the same time including more specific neural and physiological bases for the child behaviors and for the learning mechanisms. For example, would existing neural network models of learning to produce speech sounds via reinforcement, such as [35] or [36], if given realistic numbers of trials and adult reinforcement rates, also yield good fits to human data? Relatedly, future extensions of this work both on the human side and on the modeling side should take into account more detailed information about the acoustics and semantics [38] of child and adult vocalizations.…”

Sensorimotor input as a language generalisation tool: a neurorobotics model for generation and generalisation of noun-verb combinations with sensorimotor inputs

“…Then a caregiver is used to produce speech by using speech sounds for object names using reinforcement learning, where the reward is again given by the response of the caregiver. Likewise, a selforganizing map together with reinforcement learning was proposed in Warlaumont (2013), which demonstrated that the reinforcement learning based on the similarity of vocalization can improve the post-learning production of the sound of one's language.…”

The Social Feedback Hypothesis and Communicative Development in Autism Spectrum Disorder

“…Motorskill learning is influenced by reward (e.g., Buitrago, Ringer, Schulz, Dichgans, & Luft, 2004; Izawa & Shadmehr, 2011), so it is plausible that contingent social responses positively reinforce children both with and without ASD as they develop motor speech skills. Supporting this possibility, several computational models have shown how speech learning can be viewed as a type of motor learning that is influenced by reward (Howard & Messum, 2014; Miura, Yoshikawa, & Asada, 2012; Warlaumont & Finnegan, 2016; Warlaumont, Westermann, Buder, & Oller, 2013). …”