Give me iron ore and I will make you an iron. pic.twitter.com/PlJIQYVJrf— Object Lessons (@objectsobjects) April 27, 2016
Ian Bogost approached us looking to refocus the Object Lessons Twitter account and make it more of a reflection of the book series. After some discussion, we all decided that having the account itself write pitches for books would be a good use for a publishing imprint's Twitter presence.
The bot works by querying Wikipedia like a database. It traverses what Wikipedia knows about various objects in the world and attempts to make true statements or pose questions based on this knowledge.
Getting the voice and tone of a bot right is often the subtlest, hardest part of making a bot, and this was certainly true for this project. The Wikipedia ontologies we were querying are a motley data set, and not terribly internally consistent. For future projects using this kind of data, it would make sense to incorporate Tracery grammars, which would let us fine-tune the tone while maintaining the interesting data connections.