Dr Raquel Alhama (Max Planck Institute for Psycholinguistics, Nijmegen) will give a LuCiD seminar from 11-12.30 on Wednesday 12th February at the University of Manchester. She will share her research on Learning Words from Context: Predictions of Context-counting and Context-predicting Models on Vocabulary Acquisition. The seminar is free to attend (no booking required) and will take place in the Thouless room, Coupland 1 Building, University of Manchester.
Distributional learning is likely a fundamental mechanism necessary for a range of features of language. While there are well-known demonstrations that children use distributional information for language acquisition, the mechanisms underlying these achievements are unclear. In the current talk, I will present the results of a series of computational simulations that compare different types of distributional learning algorithms against vocabulary acquisition data from human children. Specifically, we compare a context-counting to a context-predicting algorithm, and show that context predicting algorithms provide the best fit to the human data.