In the runup to PyconUK 2014, I made the following ill-advised statement in an IRC channel: "I feel like I should find something to talk about at PyconUK. I wish I had something interesting to talk about." Nine seconds later someone replied "create a markov chain to generate a talk from the names of the talks at pycon and europython, then talk about how you did that, using a title it generates as the title of the talk."
Challenge accepted. This is that talk, admittedly one year late.
In this talk I will briefly describe Markov Chains as a means to simulate conversations and graph databases as a means to store Markov Chains. After this, I will discuss various considerations for creating interesting candidate responses in conversations, along with the challenges of too little and too much data. Finally, I will demonstrate my implementation and generate the title of this talk.