Quicker Bot: @AppreciationBot

Screen Shot 2014-05-20 at 16.38.43

Darius Kazemi, one of the most fun people on the Internet, launched a bot today which tweets pictures and descriptions of pieces at the Metropolitan Museum of Art. I love bots like this, and I also really like the idea of bots which riff off one another. For a while now I’ve wanted to write a bot that named colours put out by EveryColorBot, but never thought of an elegant enough trick to doing so. Today I knocked up AppreciationBot in about an hour, which tweets comments about the pieces presented by MuseumBot.

Screen Shot 2014-05-20 at 16.38.28

The bot is very basic right now but I was asked how it works so here’s the rundown. There are three components to the bot right now: a list of weasel phrases which are meaningless links between different parts of the tweet. This lets it say things like “this piece seems to hint at” something and it doesn’t really matter what comes after it – to an extent.

The real meat of the bot is in the two statements it makes in the first sentence – first, the identification of an element of the work. This is done by looking at the words used in the title of the piece, and running them past ConceptNet, specifically looking for IsA relationships. A foal IsA young mammal according to ConceptNet, so we can extract that to make it look like the bot has identified something in the piece. This will rapidly go wrong, I promise you.

The second part is the comparison to something else. You’ve heard me harp on about Metaphor Magnet enough by now so I won’t do it again, but this tool is really great. What we do here is we run the word we chose from the description into Metaphor Magnet, and we pluck out a random vehicle. MM does have a weakness in that it can extract direct equivalences at times, or get very near to them, for some kinds of term. Again, I’m sure we’ll see evidence of this as the bot continues to tweet. It inserts this after the first weasel phrase, and then it makes a value judgement on the piece, which is also weaselly because these judgements are subjective so the bot can pretty much say what it wants!

Twitter Bots are slowly becoming a point of interest in Computational Creativity – Tony Veale launched a metaphor-driven bot which uses his Metaphor Magnet, and there are tutorials and code camps planned in the next year which focus on the topic. Expect to see more soon! I’ll be putting AppreciationBot online and adding to its features if I get the time, too.

Leave a Reply