Modeling Lexical Acquisition Through Networks

We examine the nature of phonological and semantic similarity in early language learning and how the use of this information might change over the course of development. To this end, we represent the lexicon as either a phonological or semantic network and model the growth of this network. Constructing normative vocabularies from the Communicative Development Inventory norms (CDI), we utilize a preferential attachment growth algorithm to predict which words will be learned next. We quantify the extent to which the network representation improves our ability to predict which words a child will learn next. We consider the effect of the age of the child, the total words in the vocabulary and the language ability as measured through CDI percentile. Our findings suggest that the semantic representation does not outperform the baseline bag-of-words model, whereas the phonological network conditionally does. More generally, we show that the network representation influences the ability of a model to capture vocabulary growth, and we offer a method of analysis for testing representational assumptions in network models.