How do you model the mind of a book? This was the question John Higgs and I were discussing in the pub one night. John’s excellent book The KLF: Chaos, Magic, and the Band who Burned a Million Pounds was due to be released in paperback, and we wanted to see if we could make it ‘come alive’ on the internet in some way. I’d already run a generative radio station for 23 days for the initial release, Radio Eris, and wanted to do something different.
John gave me a copy of the book in it’s print-ready format, and I set about pondering how best to extract the tangled web of people and events inside it, with a view to creating some sort of interrogatable representation, that could then be shoe-horned into the internet in some way. I set about creating a program to extract the named entities in the book, all the people mentioned, and their position inside the text. From this skeleton, I looked at each entity and examined which other entities were mentioned nearby in the text. This, to my surprise, gave me a model of interconnections which bears a remarkable likeness to the ideas contained in the book.
The view below picks 3 names from the book at random – if you click on them, their relative relationships are revealed. Beware, some nodes like Bill Drummond and Jimmy Cauty are (unsurprisingly) highly connected.
Having built an index into the text, I set about attaching it to twitter, responding to questions with lines from the book. While this was quite pleasing, it didn’t quite capture the spirit of Eris who permeates the book, so I mined the text for all the adjectives and used this to form a custom dictionary of John’s descriptive style. Thus the responses of @klfbook are modified, with nouns and adjectives substituted at random inside the sentences.
John has written an excellent post about how he feels, as the author, seeing his words mangled in this way.
In addition, the book posts animated GIFs to a tumblr, based on quotes from the book. It does this at 23 minutes past midnight, 6am, mid-day and 6pm.
These GIFs are created automatically, combining a quote from the book with images harvested from google. The images are based on 23 search terms (which I’m not going to reveal here).
Much like @shardecho this project looks at how a corpus of text can behave like an intentional being, with actions, and reactions to it’s environment. Everything it does is derived from the text – the book is all it knows, and all it can ever know.