TextArc

Firstly, you choose from texts supplied by the wonderful Project Gutenberg, and then the program reproduces the words in a large oval around the screen, which looks slightly odd at first. You can't actually read the text from there. A second, concentric oval then appears inside the first one, with the text in a font that is easier to read. Inside these spirals the words appear again. Crucially, words that occur more than once are distributed inside the ellipses in their "average position" and this means that words which only appear at the beginning appear at the periphery of the large oval, and important, repeated words that appear throughout are densely packed in the middle. So, if reading Hamlet you might find Hamlet's name right in the middle. Then, what the program does is that it calculates the number of times the word appears and assigns the word a corresponding brightness, which clearly marks out key words.

Place your mouse over a word and this creates "rubber bands" that link it to every occurrence in the elliptical text, and this shows all kinds of relationships. So, patterns at the middle of the concentric ovals show significant themes, characters and their physical location in the story.

The website's rather cryptic description of TextArc says, "TextArc is a tool to aid in the discovery of patterns and concepts in arbitrary texts by leveraging a powerful and underutilized resource-human visual processing. It compliments approaches such as Statistical Natural Language Processing and Computational Linguistics by providing a gestalt overview, letting intuition help extract meaning from an unread text." I'm not entirely sure I know what that means, but the important thing is the visual representation of patterns.

It goes on to say, "words draw attention to where they appear in the document. Distribution information is revealed when the analyst points at a word: its 'rubber band' rays become visible, linking it to every place it appears in the text." This means just by rolling your mouse over a text you can see links, and this is very interesting when done with characters. You find all sorts of underlying relationships, and ideas that would never otherwise have occurred to you.

Recent additions to the program include sound: they say, "The eye's ability to pick out differences in gray (where brightness means how often a word is repeated) was a limiting factor". What they have done now is that a pitch of sound corresponds to how often a word appears, and when you move your mouse over a word the pitch of sound that you hear tells you its frequency.

I think it's a great idea because it is really the literary equivalent of mapping a graph from sets of data. Imagine that I gave people sets of data from famous curves in maths, physics or economics (the 'Philips Curve', for example). The mass of numbers would mean very little, and people would have trouble spotting the pattern or relationship - they would never get the significance.

By visually representing texts on a 'graph' of sorts we get to look at the big picture, the real meaning, the underlying patterns that are present in most works of art.