Digital Humanities

Depending on whom you ask, there are a finite number of plots in literature. This number, whether 2, 7, 20, or 36, includes plots such as “rags to riches,” “the quest,” “revenge,” “temptation,” and “deliverance.” These plots have traditionally been determined by analyzing literature and searching for common themes. Georges Polti, for example, drew from the work of Carlos Gozzi to classify 36 plots by analyzing Greek and French texts. Christopher Booker, in his book The Seven Basic Plots, used analytical psychology to describe what he believed to be the seven plots and their psychological meaning.

With a possible sample size as large as every book ever written, one major problem with identifying these plots is sample bias. Would Polti have drawn different conclusions had he analyzed Chinese texts and Native American oral stories? Is it possible that the basic plots are regional, rather than universal? Have new plots ever emerged alongside new technologies or political developments?

Recently, perhaps solving this problem, the task of classifying these basic plots has been outsourced to computers. Matthew Jockers, a University of Nebraska professor working in the field of “digital humanities,” used high-powered computers to analyze more books than any one person could read to identify six basic plots. Or seven, depending on which texts the computer selected at random. Even computers can’t solve the problem of sampling bias it would seem.

Jockers’ computers, although he has yet to release the technical details of the process, rely on sentiment changes in a story to create a classic “plot graph” of the rise and fall of events. Being a computer, such sentiment changes must necessarily be related to the physical words on the page, so one possibility is analyzing the frequency of positive and negative adjectives and adverbs to determine the overall mood of a passage. This reveals another potential flaw in the process - since a computer is analyzing the text, rather than reading, an author’s intentions may be misinterpreted.

Assigning a computer to this task raises a number of questions. Is it even possible for a computer to replace human analysis within the field of HUMAnities? Statistical analysis and data mining can only go so far when the data is being gathered by a machine that has no understanding of the data is gathering. Analyses such as these can provide valuable insight into literary analysis, to be sure, but unless the target audience, ie a human, is involved, there will invariably be error.

Another question is why does this even matter in the first place? Is this just a case of the human tendency to create classifications, even when there are none? Given the large discrepancies between the assertions (Tolstoy’s 2 to Polti’s 36), it’s fairly apparent that the determined number of plots will vary depending on how specific one wants to be. Plots can be similar in a number of ways, but it is up to the individual to decide how similar plots must be before they constitute one universal Plot.

There can be an infinite number of plots, one for each story. There can be one plot, a story in which something or nothing happens. The irony of using a computer to attempt to solve this problem is that it is a problem created by and for people – a meaningless question with no, or infinite, answers.

http://motherboard.vice.com/read/computers-find-that-there-are-six-plots