What I Thought About

[muffinista]

I gave up on my main idea for NaNoGenMo and decided to work on something simpler instead. I spent awhile exploring Google's n-gram datasets and I thought I would try generating some output from them.

I decided it might work to find a common phrase and work from there. I downloaded the 5-grams that started with the phrase "I thought about". There were a few thousand of these. I edited them down to include only complete (or mostly complete) sentences, which left just over 600 phrases.

I wrote a pretty simple ruby script to just randomly select from this output and combine into paragraphs, sometimes prefixing the phrases with other prefixes I found in the n-gram data.

The output is very repetitive and I can no longer look at the work 'thought' without hurting my brain, but I'm quite happy with the results.

Code: https://github.com/muffinista/NaNoGenMo2015/blob/master/i-thought.rb Output: https://github.com/muffinista/NaNoGenMo2015/blob/master/i-thought-output.md

[hugovk]

I showed this to my friend who recommended you get Mark Cousins to narrate it.

[auxiliary]

Very interesting results. I was looking at the last chapter and noticed a few dark ones like "I thought about killing you". It might be interesting to create emotion lists and change the tone that way as chapters progress. For example, starting with cheerful words and going towards a darker tone. Just an idea.

[muffinista]

Thanks for the feedback! I might fiddle with this for whenever generative poetry month rolls around, and that's something that is worth exploring.