Poetry Playground

So I said I was going to take a break. I did. By doing something new.

I decided to see if I could build a stack of “smart” generative word and sentence fragment tools. I say smart because it’s relatively simple to create random words and not hard to use simpler techniques like markov chains to occasionally eke out a few coherent words. What’s more difficult is to algorithmically and programmatically generate things from human-provided seeds without leveraging an LLM.

I started with just a few ideas, like:

• what if I mine Project Gutenberg texts for random sentences and just take them out of context?

  

  ---

• what interesting things can be done with calculating the levenshtein distance between two words and finding other words at the same distance?

  

  ---

• what if I took a word like “thought” and used a dictionary, but instead of looking up the word, I found every definition that has the word “thought” in it?

  

  ---

• what could I build that takes me on a word journey from a word like “hope” to “despair” in a definable number of steps?

  

  ---

• what if I combine a whole bunch of concepts? Say, take the output of one generator and send it into another with the overarching goal of sketching an idea for a poem. Not the poem itself, mind you, but just some words.

  

  ---

• what if want statistical and stylistic analysis of my own poetry?

  

  ---

  None of this uses Generative AI as we refer to it today. Just plain old math, machine learning libraries, and the datamuse API.
  
  There’s more to it than this. But this description should suffice.