Today: A Text Adventure in Space

Today: A Text Adventure in Space is a playful interface for textual expression. It is available on iPhone and iPad here.

In Today, the player navigates a word through space, colliding into new words to form a sentence. The game ends when the player's sentence collides with a period.

Words are generated using Markov chains, a simple algorithm used in natural language processing. Markov chains predict the next word based on the previous word; for example, given the word "I," a Markov chain will provide a set of possible words along with their probability of occuring after "I":

fdist("I") = ["am":0.5, "was":0.2, "did": 0.2, "love": 0.1]

These probabilities are typically based on a large source text, such as the Brown Corpus.

Today: A Text Adventure in Space creates a new way for machines and humans to jointly compose text. The set of possible words is dictated by the machine, and the path that the sentence walks is determined by the human player.

Here's the poem generated in the video above. The source text is a mashup of a chapter of Melville's Pierre: or, the Ambiguities and an original work.

Yesterday was  
        long and  
the  
universal  
sincerity  
        of his,  
    Pierre was done,  
the love  
    we touched  
        long.