My final for Reading and Writing Electronic Text is a twitter bot, named Rambling Taxidermist. While browsing the depths of Project Gutenberg early in the semester, I came across two texts that played well and/or oddly together: “The Good Housekeeping Marriage Book”, and “Practical Taxidermy”. One of my first homework assignments involved just cutting some lines out of each and making a new text. It didn’t read particularly well, but it got me thinking that they could make something funny eventually. As I thought about a final, I came back to these couple of texts. I also decided I wanted to make some kind of bot (Twitter seemed like the best option), since it was something I’d never done, and it looked fun. With the one text being a marriage guide, I envisioned a bot that would tweet marriage advice to people.
I still wanted to work the taxidermy text in, however. Learning how to pick out parts of speech from a text using Textblob presented a slightly more nuanced way to combine the texts than line splitting. I began testing out how to get nouns, verbs and adjectives out of the texts and switch them with their counterparts in the other. After a lot of experimenting, I found that taking the nouns from the taxidermy text and switching them with those in the marriage text provided the best results. But these results weren’t quite the nice little nuggets of marriage advice I had imagined; they read more like the ramblings of an insane person. My bot morphed from dispensing nice marriage advice to inane, unrelated blabbering. This idea took the form of the taxidermy fox seen above, as I pictured him searching twitter and smashing away on the keyboard in his marriage counselor office to give out crazy “advice”.
The end result is the Rambling Taxidermist. It searches twitter for people tweeting about marriage, then replies to them with a sentence from its baseline corpus, which is the marriage text with nouns replaced by those from the taxidermy text. Searching for “marriage trouble” or something similar didn’t provide enough results, even though it fit the mold of the project better. The only part that doesn’t work how I would like at the moment is due to the constraints of 140 characters. Many sentences are longer than that, and right now I’m cutting off at 120 characters (to accommodate for the @username reply). This means some tweets are just cut off mid-word, which is unfortunate. An easy solution would be to only select sentences that are under the character limit, but I feel right now that’s limiting too much of the text. I think ultimately I should search for more marriage and taxidermy texts and perform the same operations, then take the shorter sentences so I have a larger corpus to draw from. It has been enjoyable though – people have been following, favoriting, retweeting, and replying to me. The response from people has been great too, and I would like to keep refining the code and maintain this. Screenshots of some of the best tweets and conversations are below, and the code follows that.