Inferring a probabilistic model of semantic memory from word association norms
Mark W. Andrews, David Vinson, Gabriella Vigliocco · 2008
In this paper, we introduce a method of data-analysis for word association norms. The defining characteristic of this method is that is based upon the inference of a probabilistic generative model of word-associations. The inferred model can in principle provide a clear and intuitive representation of the semantic knowledge inherent in word association data, facilitate an understanding of the process by which word associations are generated, extrapolate beyond the observed data to make reasonable inferences about new word associations, and facilitate an understanding of the process underlying false recall in memory experiments. Finally, the nature and form of the probabilistic model inferred using this method is directly comparable to the so-called Topicsmodel of Griffiths, Steyvers, and Tenenbaum (2007). As such, a potential future application of this work is the analysis and validation of the semantic knowledge inferred from the distributional statistics of text by direct comparison with the semantic knowledge inherent in association norms.