These word clouds derive from a topic model trained on the dataset, with each entity treated as a "document", and each discrete observation either a word from a text field or value from a categorical label. They are colored according to the field (this may take several seconds to load).