The Blackbird Project
What does this application do?
A piece of text (a poem by Wallace Stevens, say) is entered by the user into the application. The user can then create definitions of how the words are related to each other based on what definitions they find important, grouping words that are related similarly. The application can then replace a set of related words with a set of words related similarly. Simple, no? The more relationships the user defines, the more the application has to offer, but at the same time it becomes harder to replace words that are complexly related. The singularity of the poem is achieved when the poem is so well defined that only those original words can replace themselves. When relationships are maintained somewhere below, but enticingly close to that singularity is where the magic happens.
What's the point?
Unlike other arts, text has remained something more or less created by an author by hand. Plagiarism, you say? But that is no more than using a copyrighted image without permission. There is no tool that allows an author the equivalent of taking two pictures and 'photoshopping' them together, or taking two songs and 'mashing' them up. Writing, however, has always had much to owe to previous influences, parody, and "stealing". Our application allows for (in fact, thrives off of) multiple texts being used at once. Imagine if a user entered and defined all the works of a poet like Stevens. Conceivably, they could see a "new" Stevens poem–the program will produce a poem based on one of Stevens' structures using only words that Stevens himself used. Now let us say on top of all those poems, we also added all the works of an older poet, like Chaucer (just for fun). What would we see then. We would see a "mashup" — a coherent text not only combining two different authors styles and diction, but also something like a hybrid language, a pidgin between Middle and Modern English. And if we were able to also add the works of a truly foreign language poet, like Rilke. What then? And if poetry is all that is untranslateable, is our application not translating only that - not the words, but the relationships between the words.
Do you really expect it to do all that?
At the present, I am only dealing with relationships between individual words. The framework however is entirely adaptable and scalable and should in the future be able to also work on a syllabic-level (rhyming, alliteration), a phrasal-level ('him and her' = them'), and from there to a translatable level (le : fromage :: la : baguette :: the : bread). The efficacy of the application is completely up to how clever the user defines his/her relationships.