Running Head Word Association Spaces 1 Word Association Spaces Semantic Spaces based on Free Association that Predict Memory Performance Send correspondence to
closely related to factor analysis (e.g., see Johnson & Wichern, 1998) but that can be applied to
large volumes of data.
The SVD method has been successfully applied in LSA (e.g. Landauer & Dumais, 1997)
to uncover the patterns of intercorrelations of co-occurrence statistics for words appearing in
contexts. In this research, the SVD method was applied on a large database of free association
norms collected by Nelson et al. (1999) containing norms of first associates for over 5000 words.
The result is that the words are placed in a high dimensional semantic space. Because of the SVD
method, the similarity between words in this psychological space is best expressed by the cosine
of the angle between the two corresponding vectors (see Derweester et al. 1990). We also
performed additional analyses by applying metric multidimensional scaling (MDS) methods (see
Torgeson, 1952; Schiffman, Reynolds & Young, 1981) on the norms such that the Euclidian
distance between two words in the resulting psychological space captures the dissimilarity
between two words. We will refer to the general method of placing the words in a space as well
as the space itself as Word Association Space (WAS). For an overview of all the different
methods to quantify associative/semantic similarity in this research, see Table 1.
By applying scaling methods such as SVD and metric MDS on the norms, we hope to
uncover the latent information available in the free association norms that is not directly
available by investigating simple measures for associative strengths based on the direct and
indirect associative strengths through short chains of associates (e.g., Nelson & Zhang, 2000).
The basic approach is illustrated in Figure 1. The free association norms were represented in
matrix form with the rows representing the cues and the columns representing the responses. The
entries in the matrix are filled by some measure of associative strength between cues and
热门文档