Enriching the hierarchy

Problem: Cluster hierarchies are difficult to interpret (defined in extension, rather than in intention).

We propose to add:

• Qualitative information
  key concepts and pseudo-title expressed as LAs

  e.g.,

  (java,language)

  (application,development), (development, kit), (application kit)

• Quantitative information
  measure of intra-cluster similarity (expressed as %)