coherence of clusters

Page numbers are for draft copy at present; they will be replaced with correct numbers when final book is formatted. Chapter numbers are correct and will not change now.

Clustering algorithms are usually a form of unsupervised learning possibly with cluster labels attached post-hoc. However, each implicitly or explcitly has an idea of a 'good' cluster, such as a small distance between items and the centre, or no obvious sub-clusters. For clustering of text dcuemnts, teher is often a desire to have clusters that are meaningful for humans, and various specialised coherence scores have been proposed for this.

Used on Chap. 9: page 188