Inspired partly by JM Coetzee's statistical work on Beckett's writing style, Swift-Speare is a set of experiments in machine-learning-assisted poetry composition using the Dr. Johnson prototyping framework I designed for TouchType.
I am currently compiling a collection of poetry in the style of Shakesepeare, Milton, Longfellow, and others with very distinctive styles. I am also applying these methods to novelists with distinctive styles, such as Joseph Conrad, William Gibson, Russell Hoban, and James Joyce.s
Step one: create a statistical model of the sonnets.
Step two: apply our predictive text algorithm interactively as I compose new writings in the style of famous writers.
Here is the beginning of a poem, based on a statistical model of Shakespeare's sonnets (showing the associated predictions):
Let base clouds stir the world's enshrouded tears
Which have no astronomy to be assail'd
Thus in thy fair appearance lay thy buried fears
Whose uncovering gaze my fond perception failed
> And in |
Sing sweetly of battle-borne ages
of fell foes infernal, and the king
whose woe endured hailing rage
and the death of exiled heroes.
Since nature gave the world to all our hopes
and beauty gives to us the youthful pleasures still;
No weariness I'll yield to all man's gain
Nor would I tell those powers above to rage.
The wise counsels of age shall have the spoil
of art, and elephant, and crown above
> when I have seen the |
I need somebody to be reality
The only thing that makes me crazy is your breath
and no matter what we've done whatever it could be
once in a star, once in my heart
> Dancing close, baby |
girl I swear
girl I know
yeah yeah this beat rocks