phrase nets in many eyes
Many eyes has just got a new visualisation type: phrase nets.
It's a visualisation technique to reveal hidden relationships in text, by looking at specific grammatical patterns which indicate some form of grouping, for example conjunctions like "and", "or", apostrophes etc. It's quite powerful, although it's just based on simple occurrence matching, it doesn't take real grammatical structures into account. I suspect that these visualisations could be improved slightly by adding parts-of-speech tagging to the process. They generate some pretty pictures as well, have a play with it yourself (hint: hovering over edges will display the textual context).
Here are my first experiments: