Last week in CS 547 David Glowacki gave a talk about some technology that his lab developed which allowed people to use Kinect (and incidentally LeapMotion) to view the effect that their bodies - as sort of “force imprints” on a canvas (the background) - have on particles bouncing around arbitrarily. It was really interesting seeing a serious implementation of Kinect and LeapMotion that really seemed to have legs, since I’m generally kind of skeptical of the appeal of these full-body user interface devices (almost as much as I am about voice interfaces, although my opinion on that has been changing recently as I’ve biked to work/class and wanted to know things that only Siri could tell me).

None of that is why I’m writing this post. Or not really.

His talk was largely motivated by a prominent metaphor in his field where molecules are dancers, pirouetting and waltzing around each other as they get closer and ultimately bond. This metaphor is not only pervasive in the field of genetics where scientists try to understand how medicine bonds and passes through the system, but it’s informative as well; it provides a mental model for laypeople to understand the interactions of an esoteric system. It might also prompt some novel research questions that otherwise wouldn’t have been sparked had their researchers never been exposed to totally new ways of explaining biological systems at the microscopic level, but it’s hard to say for sure.

I was thinking, though, that in the highly structured network of academia, journal articles could give some good insight into the introduction of metaphors to unrelated fields. What could we learn about new terms and metaphors based on the circumstances surrounding their introductions? I suspect something interesting would emerge from these details. Off the top of my head, you could probably look at unique words in any given journal article and see if the word graph representing any of the unique words don’t match the word graph of those words in the general word corpus.

I can give an example; if the word “pirouette” shows up a lot around the word “isomer” in some articles, but pirouette normally shows up surrounded by words like “dance”, “dressage”, “ballet”, etc… then we can infer that the word is a loaner. More importantly, we can identify these loan-words as soon as they get introduced to the lexicon of a certain field (again determined by the words used in a field, made easier by the fact that academics are nothing if not meticulous), and plot the success of a term’s adoption over time.

If this kind of approach has legs, it’d be interesting to see whether a model based on the data could predict adoption of a new term in a field. I suspect that some equation involving the term itself (the field and context of that term (if it’s borrowed from another field), its obscurity in the general population’s vocabulary, etc…) as well as the quality of the paper introducing the term (initially this would be determined by something like PageRank, a fairly conventional method for determining a paper’s impact involving the measurement of number of citations of that paper; later, one might develop a model predicting the cultural fit of a paper based on its adherence to established linguistic norms, especially within its subgroup - although my guess is that there’s no accounting for very high-impact papers).