“Imperial Voices”:

Gender and Social Class among Shakespeare's Characters, a Stylometric Approach
Jonathan Reeve

“in the imitation of these twain—who as Ulysses says, opinion crowns with an imperial voice—many are infect.” —Nestor, Troilus and Cressida, Act 1, Scene 3.

The Question

Can we quantify implicit class and gender hierarchies as represented in Shakespeare's plays?

Summary

• categorized characters from Shakespeare's plays according to occupation and social class
• extracted dialogue from those characters using TEI XML editions and Python scripting
• performed statistical analyses on the resulting text to determine stylistic similarities between character categories
• identified distinctive language patterns among these categories

Categorization of Characters

• difficult problem, since this can be very subjective
• who is considered a comic figure by one critic (or critical epoch) may be considered tragic by another
• therefore, categories were chosen based on their descriptions in the casts of characters
• categories chosen were "kings," "queens," "servants," "gentlemen," "gentlewomen," "officers," and "fools."

Choosing the Characters

• the Shakespeare character search engine at the Electronic Literature Foundation was used
• some manual editing was necessary
• there were many edge cases
• can Oberon, the king of the fairies, be considered a king for the purposes of this study?

Examples:

• Kings: Claudius (Hamlet), Lear, Oberon (Midsummer), the Henrys
• Queens: Gertrude (Hamlet), Cleopatra, Titania (Midsummer)
• Gentlemen: Roderigo (Othello), Valentine (Twelfth), many called "a gentleman."
• Gentlewomen: Margaret (Much Ado), Helena (All's Well)
• Servants: usually called "a servant" except for Launcelot, Leonardo, Balthasar, and Stefano in Merchant.
• a full list here

Data Set

• 42 plays in the Wordhoard Shakespeare Corpus
• Unadorned XML
• Note: I use this corpus uncritically, without really differentiating among those plays that are in question

Folger Library XML

    <sp xml:id="sp-0006" who="#FRANCISCO-HAM">
        <speaker xml:id="spk-0006">
            <w xml:id="w0000570">FRANCISCO</w>
        </speaker>
        <ab xml:id="ab-0006">
        <lb xml:id="lb-00051"/>
            <join type="line" xml:id="ftln-0006" 
            n="1.1.6" ana="#verse" target="#w0000580 
            #c0000590 #w0000600 #c0000610 #w0000620 
            #c0000630 #w0000640 #c0000650 #w0000660 
            #c0000670 #w0000680 #c0000690 #w0000700 
            #p0000710"/>
            <w xml:id="w0000580" n="1.1.6">You</w>
            <c xml:id="c0000590" n="1.1.6"> </c>
            <w xml:id="w0000600" n="1.1.6">come</w>
            <c xml:id="c0000610" n="1.1.6"> </c>
            <w xml:id="w0000620" n="1.1.6">most</w>
            <c xml:id="c0000630" n="1.1.6"> </c>
            <w xml:id="w0000640" n="1.1.6">carefully</w>
            <c xml:id="c0000650" n="1.1.6"> </c>
            <w xml:id="w0000660" n="1.1.6">upon</w>
            <c xml:id="c0000670" n="1.1.6"> </c>
            <w xml:id="w0000680" n="1.1.6">your</w>
            <c xml:id="c0000690" n="1.1.6"> </c>
            <w xml:id="w0000700" n="1.1.6">hour</w>
            <pc xml:id="p0000710" n="1.1.6">.</pc>
        </ab>
    </sp>

Wordhoard XML

    <sp who="Hamlet">
            <speaker>Hamlet</speaker>
            <l xml:id="sha-ham301055" n="55">
              To be, or not to be: that is the question:
            </l>
    </sp> 

Dialogue Extraction

• python script that uses the ElementTree library

• key function:

  for xml in xmls: 
  for character in characters: 
      output+=xml.xpath('.//s:sp[@who="'+character+'"]/s:l/text()')

• command-line program, built to be used in conjunction with Linux tools

  echo 'Claudius, Lear, ... ' | parse.py *.xml > kings.txt

Data Janitorial Tasks

• tokenized with The Intelligent Archive
• tokens grouped according to most frequent, given as percentage of total text

Statistical Analysis

• used principal component analysis (PCA) to find similarities among most frequently used words of each category
• compresses high-dimensional data into two dimensions, showing statistically significant vectors
• technique was pioneered by John Burrows and expanded upon by David Hoover

Problem

Problem

• probability of a given word appearing in the "kings" category depends less on style than on the number of words

Possible Solutions

• remove categories with the largest and smallest numbers of characters
• select chunks of text which represent the lowest common denominator, and randomize the selections

    for a in $(find . -name '-dialog.txt') ; 
        do cat "$a" | sort -R > "$a.random" ; 
    done

Distinctive Language Analysis

• used David Hoover's Full Spectrum Spreadsheet
• counts words that appear frequently in one category and seldom in another

Kings vs. Fools

• kings MDWs feature many proper names, including place names
• fools MDWs have less concrete words
• this suggests that fools are more likely to speak in abstractions
• fool's song from King Lear: "Have more than thou showest, / Speak less than thou knowest"

Kings vs. Officers

• most distinctive word for kings is "love," an unusual word for a soldier
• related words include "heart" (#17), "sweet" (#21), "gentle" (#32)
• kings use kinship terms: "son" (#7), "father" (#11), "cousin" (#43), "wife" (#70)
• the order of the above is perhaps telling of a kinship hierarchy

Kings vs. Queens

• kings no longer seem so sensitive in this comparison
• kings feature hierarchical words: "lords" (#2), "Earl" (#10), "Duke" (#13) and militaristic words: "sword" (#23), "march" (#39), "war" (#47)
• kings MDW kinship terms are all male: "brother" (#12), "uncle" (#20), "father" (#40)
• kinship terms among the queens' MDWs are all female: "wife" (#23), "daughter" (#24), "woman" (#27)

Gendered Pronouns

• "she" and "her" don't appear in any of the kings' MDW lists
• yet they appear high on MDW lists for fools (#4), servants (#9), and queens (#18)

Male royalty are predominantly concerned with other men, and female royalty are predominantly concerned with other women.

Non-surprise MDWs:

• fools: "ass" and "jest"
• officers: "watch," "prisoner" and "guard"
• kings: "crown"

Surprising MDWs

• gentlewomen: "power" and "blood"

A Closer Look

• a UNIX concordance:

  grep power gentlewomen-dialog.txt > power.txt
  while read p  
    do grep -B 20  "$p" *.xml 
  done < power.txt

• reveals that all uses of "power" and "blood" are from the passionate character Helena in All's Well That Ends Well

Kings: They're Affirmative

"nay" appears high on the lists for fools, queens, servants, and gentlemen, yet doesn't appear at all in kings MDW lists.

Conclusions

• to some degree, statistical analysis of character language can reveal implicit class and gender hierarchies

Caveats

• we cannot say whether these hierarchies are reflective of those hierarchies in Shakespeare's time, or in the times or places he represents
• there are many complicating variables:
• corpus selection (some of the 42 Wordhoard corpus plays have disputed authorship)
• casts of characters may not be accurate descriptors of the characters' true roles
• relative text length greatly complicates this sort of text analysis

Call for Contributions

• all of the source code for this experiment is released under the GPLv3 and available on my GitHub repository
• please add to it and improve it as you see fit

Links

• this presentation: http://jonreeve.com/presentations/cbad
• the full paper, including lists of characters and lists of most distinctive words, can be found on: http://jonreeve.com
• my email address: jon.reeve@gmail.com