Working with a small group of graduate students at the Stanford Literary Lab, English professor Franco Moretti fed a digitized text of Hamlet into a database in order to create and examine the play’s character-network:
Most recently Moretti has turned his attention to what might be the most familiar text in English literature: “Hamlet.” Using the play as a kind of test case, Moretti diagrammed and quantified the plot, charting the relationships among characters as a network based strictly on whether they speak to one another at any point in the play. He published the results in an article, “Network Theory, Plot Analysis,” in the March/April 2011 issue of the New Left Review.
Seen through Moretti’s network diagrams, “Hamlet” often seems brand new. One notices, for example, that of all the characters who speak to both Hamlet and Claudius, only two manage to survive the play (Moretti calls this part of the network the “region of death”). Or one notices that Rosencrantz and Guildenstern, the most famous pair of minor characters in all of Shakespeare, never speak to each other.
Here is a PDF version of Moretti’s article. And here is a chart of the network in Hamlet, with the “region of death” highlighted in red:
I’m intrigued by the possibilities of what could be called “algorithm-enhanced close reading” and interested in seeing the effect of such quantitative analysis on literary studies.
What do you think? Will this prove to be a useful approach to literary analysis or is it just another English department fad?
June 7th, 2011 | 10:37 am
I hereby coin a new term: “Fata” It is a portmanteau of “Fake” and “Data,” and it is most useful in describing documents like this, which purport to be chock full of useful information, but which in reality are all but meaningless.
June 7th, 2011 | 11:07 am
In his eulogy for Leo Strauss, Alan Bloom mentions that Strauss’ technique in approaching a text involved making a lot of lists, and (presumably) comparing and contrasting them. I also recall Strauss, in his Machiavelli book, doing the math on M.’s (seemingly) offhand comment that religions change two or three times every three to five thousand years, and coming up with upper and lower limits for the duration of Christianity–the lower of which suggested that its time was running out.
This is all by way of asserting that the truly great critics already approach texts as numbers crunchers–at least, that’s one tool they employ, among others. That “region of death” discovery above is interesting and potentially illuminating; but I don’t see it as very different from what great readers like Strauss and Bloom did on their own, aided only by pencil and paper.
For us non-great readers, of course, computer-aided criticism would be welcomed as a way to help us ape the patience and persistence of the truly great. But I have a feeling it will only make us lazier and less perceptive.
June 7th, 2011 | 11:28 am
I disagree with Barry Arlington. It is real data. It doesn’t seem to me to be a revolutionary way to examine a text, but it does help clarify the interactions or interrelationships between the characters.
When I read Nabokov’s Lectures on Literature I felt like I had previously been a very inattentive reader. He gleans all kinds of information from texts (for example, the floor plan of the house in Mansfield Park, the actual day of the week and date when various fictional events occur when not explicitly stated) just by paying extremly close attention to the narrative. This strikes me as somewhat similar.
June 7th, 2011 | 11:47 am
Well, as Moretti himself says, The other objection is: OK, so you found these things. So what? What does that change? I take this objection seriously. As I said, new methods need a little time to find their way, but we should have an answer to the “so what” question in a matter of a few years. Not tomorrow, but not 20 years from now, either.
Being able to visualize things in new ways does often turn up surprising connections. (Such as John Snow’s famous map of cholera cases in 1854.)
I’m not convinced it’s a great literary approach, but I don’t automatically dismiss it, either.
June 7th, 2011 | 1:32 pm
This is all by way of asserting that the truly great critics already approach texts as numbers crunchers
Not great literary critics.
Great social scientists, maybe – but that’s not the same thing at all.
June 7th, 2011 | 4:20 pm
I could certainly see this being useful in other fields, like historical theology. It would be quite interesting to put something like Aquinas’s Summa into the algorithm and treating his authorities as characters and then seeing how they interact (Mark Jordan has done something similar to this the hard way). Perhaps even more revealing would be putting several texts in and seeing how authorities change. So it could be quite useful in some forms of intellectual history.
June 7th, 2011 | 6:43 pm
BTW, “Fata” is not to be confused with the stinky goat cheese “Feta,” though I suppose Fata can also be, metaphorically speaking, stinky.
June 8th, 2011 | 8:58 am
[...] When the Algorithm Read Hamlet tells the story of English professor Franco Moretti who “fed a digitized text of Hamlet into a database in order to create and examine the play’s character-network.” Joe Carter asks: I’m intrigued by the possibilities of what could be called “algorithm-enhanced close reading” and interested in seeing the effect of such quantitative analysis on literary studies. [...]
June 8th, 2011 | 12:57 pm
The technique can be useful, but Moretti’s work often seems pretty weak. For instance, it’s true that Rosencrantz and Guildenstern never speak to each other, but only because Rosencrantz speaks to Guildenstern but not vice versa; and also because the two essentially function as a unit for most of the play — quite literally sharing speeches and sometimes even lines. Likewise it’s true that only two in the region of death survive, but there are only nine in the region of death in the first place (if one includes Claudius and Hamlet themselves), the bulk of whom are the major characters in a tragedy and thus (a) get a fair amount of stage time with other characters and (b) are likely to die. If it were all of them, that would be something to begin with. But if there are exceptions, it is useless information unless one has a good reason why those exceptions have to be there. (It would be easy with Horatio who, despite not being the main character, is the character who actually unifies the play as a whole, but Osric? If some really good reason, supportable by textual evidence, were forthcoming, it would be one thing; otherwise, it’s seeing patterns that probably aren’t really there.)
Although he doesn’t discuss this particular venture, Cosma Shalizi has a good discussion of Moretti’s overall approach (in the context of his work on genre):
http://www.thevalve.org/go/valve/article/graphs_trees_materialism_fishing/
June 8th, 2011 | 5:03 pm
I think it’s pretty cool!
June 9th, 2011 | 6:37 pm
I think its important to read for both details (Nabokov’s observances in Lectures in Literature) but also for big picture content. The fun part is in the relationship between the two and how artfully and creatively, subtly or allegorically that is managed. What I worry is that by reducing particular texts and details into algorithms, readers will forget to or be unable to look past the details into the actual meaning. I’m all for details and lists and charts but only as a tool to reach the larger purpose of a text. You have to have the basic plot down before you can move deeper into a novel. Details are simply clues to characters, ideas, and meanings.
June 10th, 2011 | 12:20 pm
[...] this week I mentioned a project that performed quantitative analysis on Hamlet and wondered whether it would prove to be a useful approach to literary analysis. Last night I [...]
June 11th, 2011 | 11:42 pm
Post-modern art… can’t you see it – an entire show of the Algorithmic thumbprints of all of Shakespeare’s plays!
Links
Blogs
Find Us
Contact