For this assignment, we had to create a text using at least two source texts and the list variable type. You can see all the source files in my GitHub repository

Sources

In an attempt to join my exploration of authors and topics, I chose some works by the poet Eileen Myles. I read her autobiographical book “Chelsea Girls” last year, and I loved her writing. I used her poems “Peanut Butter” and “Each Defeat”. At the same time, I used Emmeline Pankhurst’s speech “I Incite this Meeting to Rebellion” as a bank of words.

Poem in paperspace

First, I wanted to create a poem where the form of the text was important. Wanting to replicate patterns that have formed while doing work in the terminal, I wanted to create on purpose a flow with the structure of the text.

The first stage was creating alternating lines of source material and made up lines with a fixed length, while indenting each line more and more. Each made up line uses words from the word bank, but only the ones that appear four times or less. Also, I had each line begin with the first letter of the previous one. This created the text I liked the most out of these experiments (which had an error in the code, as it read the first character of the previous line, which was a space (‘ ‘)):

I am always hungry
  incitement importa
    sort. Pleasure eye

        a woman with early

            a woman with other

Afterwards, I corrected this mistake and added the possibility of incorporating the overflowing text (result of cutting the lines to a certain length) into the next line.

Even though the code kept getting more and more complex, the results did not please me at all. It took me a long time to have an idea of what to do, but having it turn out so bad was kind of frustrating. I wanted to make the text flow and maybe my ideas should do that too.

(This first jupyter notebook can be found here)

Flowing poems

This time I took a more “musical” approach. I would select words according to the letter they began with. I made three lists of beginning letters:

l_strong = ['b', 'c', 'p', 't']
l_flow   = ['d', 'f', 'g', 'l', 's']
l_vowel  = ['a', 'e', 'i', 'o', 'u']

And then I started mixing lines of poems with each of these words, always accompanied by a word starting with a vowel. I mixed a lot of these experiments, with one or two made-up lines, three poem lines together, etc. In the end, I got something I am satisfied with (finally)

one day was
further earnest So whore in
my death. Not
prison opposition
  There
  Take
people Ulster governments evidence disregard
let me move


like a playground
far Ulster go Unlicensed
ride.
teach important my child
  Commons
  Chartists
than And sacrifice other dared
the mail.

(This jupyter notebook can be found here)

Afterthoughts

I feel like I have to learn about natural language processing in order to make stuff that at least makes some sense (or stuff that I could build some sense into). I started using the nltk library, but I need to get more into that to use it correctly (specifically the tagging… It failed when I tried to use the pos_tag() function).

At the same time, I feel I just need to experiment more in order to have more ideas and a sense of what and how I can create better stuff.