Comedy and Tragedy in Shakespeare

Project page for A Midsummer Night's Project: Comedy and Tragedy in Shakespeare

"O any thing, of nothing first create! O heavy lightness! serious vanity! Misshapen chaos of well-seeming forms! " -- Romeo on complex systems

"And thou art but of low degree." --Iago to Cassio in Othello (he's wrong)

HASTINGS: But, by your leave, it never yet did hurt

To lay down likelihoods and forms of hope.

LORD BARDOLPH: Yes, if this present quality of war,

Indeed the instant action: a cause on foot

Lives so in hope as in an early spring

We see the appearing buds; which to prove fruit,

Hope gives not so much warrant as despair

That frosts will bite them. When we mean to build,

We first survey the plot, then draw the model;

And when we see the figure of the house,

Then must we rate the cost of the erection;

Which if we find outweighs ability,

What do we then but draw anew the model

In fewer offices, or at last desist

To build at all? Much more, in this great work,

Which is almost to pluck a kingdom down

And set another up, should we survey

The plot of situation and the model,

Consent upon a sure foundation,

Question surveyors, know our own estate,

How able such a work to undergo,

To weigh against his opposite; or else

We fortify in paper and in figures,

Using the names of men instead of men:

Like one that draws the model of a house

Beyond his power to build it; who, half through,

Gives o'er and leaves his part-created cost

A naked subject to the weeping clouds

And waste for churlish winter's tyranny.

King Henry IV, part II: I, iii

Previous project description here: [1].

And we classify:

With an auspicious and a dropping eye,

With mirth in funeral and with dirge in marriage,

In equal scale weighing delight and dole

King Claudius, Hamlet: I, ii

Data and code currently in the Dropbox folder.

Notes on what is in Dropbox

For those who don't want to use the xml (shakespeare.xml in the Dropbox): in the Dropbox is a folder plain_text that has all the plays in plain text form. Act.Scene numbers are in hashtags like #1.1# on their own line and the name of the speaker is in hashtags before their speech. The words and punctuation are all separated by spaces.

The folder edge_lists contains the edges in the network with weights (currently how many times each pair of characters in each play talks to each other).

load_pickles.py is set up to load pickled (yodelay hee hoo!) objects that have been created. For instance, if you run load_pickles.py, you will get an object called networks which contains weighted networks for each play.

network.py can be used to generate networks by defining a function to calculate networks within a scene (see interactions() for a model) and then passing that function name as an argument to network.py. It outputs data into networks (edge lists, pickled network object, and plots)

character_counts.txt contains title, play, and number of speeches, lines, and words spoken by each character.

Text adjustments

Changed "Gloucetser" to King Richard III so that it's consistent across play.

Related work and links

Papers

The small world of Shakespeare’s plays

James Stiller, Daniel Nettle, Robin I. M. Dunbar

Human Nature, December 2003, Volume 14, Issue 4, pp 397-408

This contains both an early network science spin (note "small worlds" in the title and publication year) as well as an evolutionary psychology/biology spin, particularly focusing on group size. A key note here is that one of the authors is Robin Dunbar, of Dunbar's number [2]. Again, this is network analysis circa early 2003, so they're really interested in connectance, average path length, clustering coefficient, and they use undirected co-occurrence within scene to determine the links. The paper is available in the Dropbox and here [3].

Hughes, James M., Nicholas J. Foti, David C. Krakauer, and Daniel N. Rockmore. "Quantitative patterns of stylistic influence in the evolution of literature." Proceedings of the National Academy of Sciences 109, no. 20 (2012): 7682-7686. This paper is available here [4]

Miranda, Pedro J., Murilo S. Baptista, and Sandro E. de S. Pinto. "Analysis of communities in a mythological social network." arXiv preprint arXiv:1306.2537 (2013). [5] (written about here [6]) Abstract: The intriguing nature of classical Homeric narratives has always fascinated the occidental culture contributing to philosophy, history, mythology and straight forwardly to literature. However what would be so intriguing about Homer's narratives' At a first gaze we shall recognize the very literal appeal and aesthetic pleasure presented on every page across Homer's chants in Odyssey and rhapsodies in Iliad. Secondly we may perceive a biased aspect of its stories contents, varying from real-historical to fictional-mythological. To encompass this glance, there are some new archeological finding that supports historicity of some events described within Iliad, and consequently to Odyssey. Considering these observations and using complex network theory concepts, we managed to built and analyze a social network gathered across the classical epic, Odyssey of Homer. Longing for further understanding, topological quantities were collected in order to classify its social network qualitatively into real or fictional. It turns out that most of the found properties belong to real social networks besides assortativity and giant component's size. In order to test the network's possibilities to be real, we removed some mythological members that could imprint a fictional aspect on the network. Carrying on this maneuver the modified social network resulted on assortative mixing and reduction of the giant component, as expected for real social networks. Overall we observe that Odyssey might be an amalgam of fictional elements plus real based human relations, which corroborates other author's findings for Iliad and archeological evidences.

From Moretti's Lit Lab at Stanford: "Network theory, plot analysis" [7] and "Quantitative Formalism: An Experiment" [8]. The first pamphlet (also appears in the New Left Review in a slightly different form [9]) considers the network based on speaking events in at least Hamlet, Macbeth, Lear. Fascinating writing style about network analysis.

Agarwal, Apoorv, Augusto Corvalan, Jacob Jensen, and Owen Rambow. "Social Network Analysis of Alice in Wonderland." NAACL-HLT 2012 (2012): 88. [10] uses hand-coded (!) social events in Alice in Wonderland.

Graham Sack. "Character Networks for Narrative Generation" AAAI Publications, Eighth Artificial Intelligence and Interactive Digital Entertainment Conference [11] Abstract: In this position paper, the author proposes the use of social networks of characters as an AI narrative generation mechanism. The first part of the paper offers examples of recent research by literary critics on the relationship between character networks and narrative structure. The second part of the paper offers a simple example of story generation based on a structural balance network model.

Graham Sack. "Simulating Plot: Towards a Generative Model of Narrative Structure" [12] 2011 AAAI Fall Symposium Series > Complex Adaptive Systems: Energy, Information, and Intelligence. Abstract: This paper explores the application of computer simulation techniques to the fields of literary studies and narratology by developing a model for plot structure and characterization. Using a corpus of 19th Century British novels as a case study, the author begins with a descriptive quantitative analysis of character names, developing a set of stylized facts about the way narratives allocate attention to their characters. The author shows that narrative attention in many novels appears to follow a “long tail” distribution.The author then constructs an explanatory model in NetLogo, demonstrating that basic assumptions about plot structure are sufficient to generate output consistent with the real novels in the corpus.

R. Alberich, J. Miro-Julia, F. Rossello, Marvel Universe looks almost like a real social network (February 2002) arxiv:0202174 [13] They look at comic book co-occurrence of characters in the Marvel Universe. Again, note the date on this - they show that this network is not like an ER-like random graph, but has (eg) too low of a clustering coefficient to be like a real-world social network.

Learning Latent Personae of Film Characters. [14] David Bamman, Brendan O’Connor, and Noah A. Smith. ACL 2013, Sofia, Bulgaria, August 2013 Abstract: We present two latent variable models for learning character types, or personas, in film, in which a persona is defined as a set of mixtures over latent lexical classes. These lexical classes capture the stereotypical actions of which a character is the agent and patient, as well as attributes by which they are described. As the first attempt to solve this problem explicitly, we also present a new dataset for the text-driven analysis of film, along with a benchmark testbed to help drive future work in this area.

with very good blog post here: [15]

Cosma's notebook on "Joint modeling of texts and networks" [16]

Blog posts, etc.

There have been a few informal explorations of this so far that I've found.

There are a set of animations of the dynamics of communication in Shakespeare plays [17] , using PieSpy [18], a heuristic-based tool to infer social networks from IRC (PieSpy paper here [19]). There isn't any analysis here and the data isn't available, but we could check out PieSpy's rules for inferring relationships. (Our methods so far: co-occurrence and before- and after- speaking turns; easy next steps could use weighted co-occurrence--as in the next set of blog posts--and longer-memory speaking turns.)

Prosody & social networks in Shakespeare: [20]. This class does a few projects on social networks of "frequency of communication" between characters. They have a cool (but not pretty) visualization of communication links between characters, colored by type of communication - verse or prose.

Wordseer has a number of explorations of language in Shakespeare plays: on men and women [21] and on "beauty" [22]. They look at grammatical relationships, word order, word frequency by type of play, collections of words, etc. They do a preliminary look at gender and love in Shakespeare. (A nice sample result: "A picture emerges: women’s most commonly-mentioned possessions are their male relatives and their bodies.")

Set of blog posts using co-occurrence to construct social networks in Shakespeare, starting here: [23]. These are co-occurrence networks, where co-occurrences is measured within scenes, and the networks are broken down by acts. Edges are weighted (and sometimes thresholded) by the number of scenes with dyad co-occurrence. They point out that you can start to see the plot of the Tempest, though I think our networks showed that more clearly.

The second blog post of the set, comparing social network density (E/V^2) in tragedies and comedies: [24]. They make a cute observation that these networks are highly connected in Act 5 when the play is a comedy, which often end in weddings (and thus high co-occurrence).

Some other fun places to look: titles at the humanities workshop at NetSci, a networks conference held this week [25]; the Six Degrees of Francis Bacon [26].

Also: xkcd does character interactions over time in Lord of the Rings, Star Wars, Jurassic Park [27]; social network analysis in Lord of the Rings prezi [28]. In the latter (by Luke Hillman [29]), he tries the Bechdel test [30]. (I'd like to try this; I'll check again if that's in the Wordseer blog post. - azj)

Also to be explored more: the Franco Moretti's group, the Stanford LitLab [31]; MovieGlaxies movie social networks (paper) [32]; Reading Maps Graphs Trees (responses to Moretti) [33]. - azj

Possible and previously discussed project directions

A non-exclusive, non-exhaustive and otherwise rough list of ideas that have come up recently. We could use this page to structure some of our results or sub-projects (see next section).

Predicting tragedy/comedy(/history)

Gender

Gender and character roles

Dynamics of the network(s)

Probabilistic modeling of the networks (consider also tragedy/comedy, gender, dynamics and evolution, etc.)

Language in Shakespeare (consider also tragedy, gender, dynamics, etc.) - more here.

Hierarchical/social roles in Shakespeare

Story structure in Shakespeare

Narrative style and structure (focus on main characters vs. more contextual/observed scenes)

etc.

Maybe try building up our own networks where, each time step, a character is either born and talks to someone or an existing character talks to another existing character. There would be parameters for how likely characters are to appear and how likely they are to interact with other characters, and we could vary these to generate comedies and tragedies? I dunno. KM

Possible uses for this page

Structuring or collecting results

Collecting information about the data or code

Sharing analyses

Nothing after this week

Showing off our results at the end of the summer school

etc.

Linear Mixed Effects Model

I've Dropboxed prop_frame.Rdata, adj_mat.RData and shakes_v1_1.R to run some simple analyses on network dynamics. You can call interaction matrices of each play in each act from the list of all plays using adj.mat[[i]] where i is a number from 1 to 135. The data frame, props calls a dataset of network properties (simple properties for now) of length 135. The play and act in row i corresponds to the weighted adjacency matrix adj.mat[[i]].

-Ashkaan

Types of Network Construction

To be included in make_network_of_scene in network.py

Types/wishlist:

Co-occurrence within scene (i <->j if i and j appear in scene s)

Speaker/listener (i -> j if i speaks before j), i.e., directed before/after

Before/after speaking (number of times, i <-> j if i speaks before j)

Binary before/after (i <-> j if i speaks before j)

Before/after speaking, longer memory (i <-> j if i speaks one or two or m turns before j)

Generative Drama

Comedy / Tragedy / History

Generate network structure through time.

(1) Generate a time series of speeches. Each point is a speech given by a specific character.

(1a) Generate a new character at anytime.

(1b) Each character that exists has a conditional probability of speaking after all the other characters. E.g., P(Romeo | Juliet)

(1c) The above conditional probability can be decomposed into how often that character talks. E.g., P(Romeo | Juliet) = P(Juliet | Romeo) * P(Romeo) / P(Juliet)

Full Model

And his love Thisbe; very tragical mirth.' Merry and tragical! tedious and brief! That is, hot ice and wondrous strange snow. How shall we find the concord of this discord?

PHILOSTRATE: A play there is, my lord, some ten words long, Which is as brief as I have known a play; But by ten words, my lord, it is too long, Which makes it tedious; for in all the play There is not one word apt, one player fitted: And tragical, my noble lord, it is; For Pyramus therein doth kill himself. Which, when I saw rehearsed, I must confess, Made mine eyes water; but more merry tears The passion of loud laughter never shed.

Yea, this man's brow, like to a title-leaf, Foretells the nature of a tragic volume: