This research represents an attempt to model vocabulary acquisition in children. A computational model, based on Feigenbaum and Simon's EPAM theory of perception and learning, is being developed. The intention is to model both how new words are acquired and the relative distributions of categories of words acquired.
Vocabulary acquisition in children begins slowly, yet after learning 40-50 words there is a vocabulary spurt. In addition, the distribution of categories of words learned changes over time. This research will use EPAM to explain how vocabulary is acquired, why there should be a sudden spurt in vocabulary learning, and why the categories of words learned changes over time.
Gathercole and Baddeley (1989) have used the phonological loop to explain vocabulary acquisition. However, they do not specify how the loop interacts with long-term memory. Our current research has implemented the phonological loop within the EPAM architecture, and proposes a method by which the loop can interact with long-term memory. The input to this model is from a large-scale naturalistic study of children's early grammatical development, allowing the influence of the input to the model upon vocabulary acquisition to be examined (see Jones, Gobet & Pine, 1999). The words in the input are converted to phonemes using the CMU Lexicon database. A suite of files to access the lexicon, and convert utterances to phonemes, have been developed and are freely available. (On a Unix machine the file can be expanded by first entering "gunzip lexicon.tar.gz" and then "tar xvf lexicon.tar".)
Our next step is to examine the types of words that the model learns first, and whether the model shows a similar vocabulary spurt to that seen in children.
This is a PERSONAL page and does NOT necessarily reflect the position of Brunel University
Last Modified: 04/10/2004