Description
I run a Twitter bot that's aimed to be amusing. It generates tweets based off a mixture of other Twitter feeds, Facebook posts and some other sources. One of the challenges in making this work has been getting decently amusing output.
This presentation will go through some of the options available, ranging from simple Markov systems using markovify, through more complicated Markov approaches and will then look at systems based on torch-rnn and other neural network systems.
You'll leave this presentation with a decent overview of the ways you can transform content into new content and the strengths and weaknesses of each.