Computational modeling of early language acquisition: what and why? Adjunct Professor, Research Fellow Okko Räsänen (Aalto University)

Computational modeling of early language acquisition (LA) refers to the use of mathematical models and computational simulations to understand language learning in infants. The basic idea is to build functional models that are capable of learning linguistic patterns and representations from similar sensory and interaction data that language learning infants are exposed to, and to replicate human learning patterns from similar language experience. This allows us to ask what kind of language structures are learnable from certain input, what kind of a priori learning mechanisms or constraints might be needed in the process, and how different factors such as quality and quantity of language input or infant-caregiver interaction strategies may impact language learning outcomes. The purpose of this talk is to provide a high-level introduction to the modeling of LA, briefly describing what it means in theory and practice, why it is justified based on what we know about learning and cognition, and how it relates to other research methods in language learning. I will also argue why computational modeling is invaluable in the process of putting together the bits and pieces of empirical findings and theoretical considerations into a coherent big picture of language learning.