Users can drill multiple levels down. For example, we select the cluster sippy cup, then browse its sub-clusters.
For ultimate flexibility users can define a subset of documents to be analyzed via a search query or a text classifier.