Nonnegative Matrix and Tensor Factorization for Discussion Tracking - Text Mining: Classification, Clustering, and Applications - page 112

judgment to the quality of the clustering and can often attach a label as well.
For instance, the cluster over which the yellow crosshairs of Figure 5.9 lie
also contains the terms (among others) football , longhorn , Texas , quarterback ,
score , redshirt , freshmen , punt ,and tackle , prompting a user to potentially
label this cluster Texas Longhorn Football .
The vismatrix tool also allows a user to quickly scan document labels as
well. Thus, hidden patterns that pertain to the documents can be found.
For instance, this Enron dataset contains one small cluster of 12 documents
using 447 terms. Figure 5.10 is a close-up 6 of this part of the reordered Enron
term-by-email matrix.
FIGURE 5.10 (SEE COLOR INSERT FOLLOWING PAGE 130.) :
Close-up of one section of pixel plot of the reordered Enron term-by-document
matrix.
6 The vismatrix tool also contains zoom in and zoom out features.