This Twitter bot generates a metaphor every two minutes (in spite of its name, since Twitter places limits on automated posting), and it is more than sufficient. The constraint provides a little breathing room to consider the metaphor before facing a new one. How does one approach this steady stream of conceptually challenging metaphors?
According to George Lakoff and Mark Johnson’s cognitive linguistic metaphor theory, a metaphor is a mental process of thinking of one conceptual domain in terms of another. Their method to study metaphor was to identify the source and target domains and map the linguistic expression of the metaphor across both. The poetic practice of the conceit or extended metaphor lends itself well to this kind of analysis, because it overtly explores the connection across both.
Kazemi’s generator produces metaphors in conceptual structures that lend themselves to this kind of mapping.
The sentence structure in each is the same, and it is best described by Kazemi himself (as he does in a fascinating blog entry about this generator): “(a/an) noun (is/considers/of) (a/an) noun: adjective (and / , not / , yet / but / , / , but not) adjective.”
Most of the generated metaphors are like the second example above
(“a slipknot”), which establishes the metaphor in the first clause and describes it in the second. So the metaphor invites us to think of the slipknot in terms of the trilogy, describing it as something without merit and monastic. The three part structure of the trilogy helps us understand the ascetic simplicity of this rope tying technique, but why is it meritless?
Normally, we could approach the question by thinking about what the person who created the metaphor may have intended by it. In the case of a more elaborate and constrained generative work, we might get some insight by finding patterns of meaning in the variations or the dataset, or even reading the source code to examine documentation, variable names, and dataset contents. However, this work is built from a very brief and basic formula that connects to a huge dataset, the Wordnik library, so we can only attribute intention to Kazemi as far as the creation of these kinds of metaphoric expressions are concerned.
But there’s a twist. Kazemi inserted a clever mechanism that allows human intention to adhere closely to some of these metaphors.
The only other interesting thing the bot does from a programming perspective is that every five hours, it grabs the last 20 retweets of its own tweets and favorites them. The result is that you can see a best-of list right here.
The two examples cited above were taken from that “favorites” list, which means that someone found them compelling enough to retweet. Here’s the slipknot metaphor with expanded information in Twitter:
We could go on to explore “roux ルー {hiatus}” and try to determine what that person could’ve found intriguing about this metaphor, but we don’t really want to be overly deterministic and fall into an intentional fallacy. Suffice it to say, we have a human connection on a work that has been curated (in the most basic use of the term).
What Kazemi has created is a minimalist little bot that produces a stream of structured language with the potential to capture our attention and provoke us into thought about conceptual relations we might not have otherwise established. And by connecting it to a social network’s functionality, he has allowed us to share our delight in these metaphorical thought-generation machines.
Featured in Genre: Bot