BotCache

CapsCop finds users who have tweeted in all caps and contacts them to tell them to stop.

Every minute, and sometimes several times a minute, CapsCop identifies a user who has tweeted using all capital letters. The bot replies to the user’s tweet with one of several witty, preprogrammed responses. CapsCop also adds users to several Twitter lists with names like “Jail” and “Life Imprisonment” if they are contacted by the bot frequently.

CapsCop seems to be viewed by some as a troll, others as a public service, and still others as a source of entertainment. Users who have been contacted by the bot sometimes retweet the bot’s message to their followers. Others reply to the bot with a comeback joke, a simple “thank you,” or an irate message for intruding into their conversations. Less frequently, users familiar with the bot who notice a friend tweet in all capitals in their twitter feed will invite the bot to contact their friend by mentioning the bot (e.g. “cc @capscop”).

The bot’s creator keeps track of user statistics and explains some additional features on capscop.com. Contact the bot with another user’s Twitter handle and the hashtags #snitch or #narc, and the bot will search the user’s recent updates specifically, looking for an all-caps tweet to trigger a reply. Those who mention the bot with the #stats hashtag will receive a reply with an all-time count of times contacted. Finally, while it’s not mentioned in the bot’s user profile or posted on the capscop.com site, users can opt out of potential future interactions with the bot by simply mentioning the bot and a safe word, “banana.”

Horse_ebooks tweets excerpts and randomly generated lines from ebooks as spam. Although the bot’s creator probably intended to launch the bot to drive traffic to an online bookstore through affiliate links, the bot’s incomplete, ungrammatical non-sequiturs are appreciated by followers as forms of computer-generated poetry, aphorisms, or expressions of emotion.

It works by taking text from a large library of ebooks (linked in the bot’s user profile) and randomly generates short statements, probably with a Markov text generator. This text is most likely tweeted as a tactic to evade Twitter spam detection. Every fifth tweet, the text is followed by a link to a Clickbank-based ebook store. Originally, the bot tweeted through a custom application, but in September 2011, the bot’s algorithm was updated to tweet using the web interface, causing suspicion that it was no longer machine-operated by its fans.

Most interaction with the bot by fans are through retweets and favorites. Many users also reply to the bot: they finish its sentences, create derivative jokes from its tweets, or respond to the bot’s statements as if the bot were a person. Others have taken the bot’s tweets and made derivative art for fans of the bot. Some examples: writing a fanfic using the bot's tweets as elements of the story, drawing a comic to tell the story of a tweet, retweet or post the bot’s tweets in a poetic sequence, and creating a web browser plugin to replace a web page’s text with the bot’s tweets.

Further reading in Something Awful, Title Magazine, Gawker, Splitsider, Time, Know Your Meme, Mr. Destructo, The Daily Dot

BotCache co-founder Alex Leavitt spoke today at SXSW on a panel titled Bots for Civic Engagement, alongside Erhardt Graeff, Greg Marra and David Bausola.  MIT Media Lab researcher Catherine D'Ignazio wrote a summary, which includes a mention of BotCache:

Are we experiencing emotional connections? Alex speaks about botcache.com. There they are tracking Twitter bots to see what connections they are making with followers and followees. One well-known account is @stealthmountain. People see it as just another person reading their tweets unless they actually check the profile. If they only interact with it on having an @ mention, you’ll get lots of things like “I didn’t realise I did that, thanks for pointing it out.” We also see people cursing at the bot, “You’re such a grammar Nazi.”

Stealth Mountain tells users who have tweeted “sneak peak” that they have used the expression improperly, and suggests that they meant to write “sneak peek.”

Every two minutes, the bot searches for the most recent tweet containing the phrase “sneak peak,” and replies to the tweet with “I think you mean ‘sneak peek’”.

The main audience for the bot is the users it contacts. Many users, when notified of their error, reply to the bot, and the nature of the replies varies. The most common replies are a simple “thank you”, excuses for making the error, enraged attacks or dismissals, and laughter.

Though the bot has over 20,000 followers, its only tweets are @-replies to users, so followers will only see the bot’s tweet if it contacts them or a user whom they follow.

Pentametron creates poetry from the natural language of other users. Random couplets of rhyming tweets in iambic pentameter – about anything – are retweeted by Pentametron to form a twitter feed in the style of successive Shakespearean sonnets.

An example stanza:

ENVIRONMENTAL SCIENCE ESSAY DONE

Cheese Macaroni *offers* anyone??

What's better? Bacon or the Internet?

That was the fastest half an hour yet

Creator Ranjit Bhatnagar designed an algorithm to analyze the rhythm of tweets flowing through Twitter’s streaming API using CMU’s pronouncing dictionary. The algorithm will retweet a random tweet from the stream if it has ten syllables in iambic rhythm (da-DUM da-DUM). Every other time, the tweet must also rhyme with the tweet before it to create an “AABB” rhyme scheme. The algorithm usually finds a matching tweet every five minutes, generating a continuous stream of couplets that are composed into sonnets on pentametron.com.

Alone, Pentametron’s retweets are simply random, concise, English statements. Together, they are poetry. Thus, fans of the bot follow the bot to read each line, and users do not often retweet the bot’s compositions. Few of the random users who are retweeted to create the poetry acknowledge the bot or recognize that their words have been added to a poem.