The results of singular value decomposition of
a matrix of text profiles are displayed in two or
three dimensions.
Note that for easier viewing, configurations are rescaled by
their
largest absolute coordinate value:
Consider
a
document vector passing through a configuration of stimulus points with
orthogonal lines drawn from the points on to it. It is the values
given to the points at which these orthogonal lines meet the vector
which
are maximally correlated with the data for that document. The document
vectors are
normalised (for convenience only) to the same length, so that their
ends
lie at a common distance from the origin of the space, forming a
circle, or
sphere, according to the number of dimensions plotted.