5 - Text Data Mining with Minimal Spanning Trees

Abstract

Publisher Summary This chapter presents a new method for effective exploration of a labeled interpoint distance matrix. The methodology supports the discovery of serendipitous relationships between the various categories encoded within the matrix. It also supports cluster structure exploration. The chapter discusses present work in the application of graph theoretic techniques to facilitate the data mining process. To illustrate the success of the formulated approaches, it provides results based on mining two small sets of text documents emphasizing that the graph theoretic approaches and visualization frameworks are not necessarily predicated on the data type being text data, but are equally appropriate for other data types. In fact, one must merely be able to compute some sort of interpoint distance measure between the observations. This interpoint distance measure does, of course, define a complete graph on the set of vertices (observations).