Home Paul Vogt
More about me

    Vogt, P. and A. D. M. Smith (2004) Quantifying lexicon acquisition under uncertainty In Lenaerts, T., Nowe, A. and Steenhout, K. (Eds.) Proceedings of the Annual Machine Learning Conference of Belgium and The Netherlands (Benelearn) 2004.

    Abstract This paper presents a mathematical model for predicting the time learners need to learn a lexicon of a given size and a given level of uncertainty. The model is based on a computational model implementing a {\em cross-situational statistical learner}. We show how well the mathematical model compares to our computational model, and compare it with another cross-situational learning model presented by Siskind. Finally, we show how we can extend our model, which was initially based on a uniform distribution of word frequencies, to a model that assumes a distribution according to Zipf's law. This, in addition, allows us to compare our model with findings from psycholinguistics.