gistfile1.py

import networkx as nx
from lxml import etree
import re
import itertools

def getNamesInAction(action,textNames,nameDict):
    # go through the names, in order of length, get them from the action, then remove them before looping
    act = action
    sortNames = sorted(textNames, key=len, reverse=True)
    returnNames = []
    for nm in sortNames:
        if nm in act:
            returnNames.append(nameDict[nm])
            act = act.replace(nm," ")
    return returnNames

# obviously you point to the .htm file wherever you have it
# find the full plays at:
# http://shakespeare.mit.edu/
filename = "ShakespearGraphs/Antony and Cleopatra  Entire Play.htm"
f = open(filename)
html = etree.HTML(f.read())
f.close()
# get all the characters in the play
chars = html.xpath("//b/text()")
# go through the list and make a mapping of the various spellings to the correct spelling we'll use for the graph
inTextToActual = {}
for i,c in enumerate(chars):
    c2 = re.sub("\s\s+"," ",c)
    c2 = c2.title()
    inTextToActual[c] = c2
# the character names as we want them
charNames =list( set(inTextToActual.values()))
# the character names as they are in the text
charTextNames = list(set(inTextToActual.keys()))
# the italics elements contain all of the entrances and exits (and other things)
# go through the text and get all the actions, and the speakers that the action is associated with
italics = html.xpath("//i")
speaker = []
action = []
for i in italics:
    ac = i.findtext(".")
    try:
        # the general stucture of the play is:
        # <a>
        #    <b> <name of speaker> </b>
        # </a>
        # <blockquote>
        #    ....
        #    <p>
        #       <i> <the action that happens> </i>
        #    </p>
        # </blockquote>
        # so we back up from the italics to get the name of the speaker
        # this lets us associate an "exit" to a character name
        sp = i.getparent().getparent().getprevious().xpath("b/text()")[0]
        speaker.append(sp)
    except:
        # in case there isn't a speaker or there is an error
        speaker.append(" ")
    action.append(ac)
# make a list of the groups of characters that are on stage together at the same time
# start with an empty list
charactersTogether = [[]]
ind = 0
# loop through actions and add characters to the set
# whenever a character is removed or added, we make a new set
for ac,sp in zip(action,speaker):
    firstWord = re.search("(Enter|enter|Exeunt|Exit|exeunt|exit)",ac)
    if firstWord is None:
        continue
    #print ac#,"\n\t",
    firstWord = firstWord.group()
    if firstWord.lower() == "enter":
        # figure out who entered
        entrants = getNamesInAction(ac,charTextNames,inTextToActual)
        if entrants == []:
            #print ""
            continue
        #print firstWord, "\t", entrants
        prevList = list(charactersTogether[ind])
        prevList.extend(entrants)
        charactersTogether.append(prevList)
        ind += 1
    elif firstWord.lower() == "exit":
        # then the character exists
        # it could be the speaking character, or not
        leavers = getNamesInAction(ac,charTextNames,inTextToActual)
        if leavers == []:
            #print "Exit \t",[sp]
            leavers = [inTextToActual[sp]]
        #else:
        #    print "Exit \t",leavers
        prevChars = list(charactersTogether[ind])
        # remove the characters
        for c in leavers:
            try:
                prevChars.remove(c)
            except:
                continue
        charactersTogether.append(prevChars)
        ind += 1
    elif firstWord.lower() == "exeunt":
        # figure out if everyone leaves or just some characters
        if ac.lower().strip() == "exeunt":
            # then everyone leaves
            #print "Everyone exits"
            charactersTogether.append([])
            ind += 1
        elif "all but" in ac.lower():
            # then everyone but some characters leave
            nonLeavers = getNamesInAction(ac,charTextNames,inTextToActual)
            charactersTogether.append(nonLeavers)
            ind += 1
        else:
            # then the characters in the list leave
            leavers = getNamesInAction(ac,charTextNames,inTextToActual)
            #print "Exit \t", leavers
            prevChars = list(charactersTogether[ind])
            # remove the characters
            for c in leavers:
                try:
                    prevChars.remove(c)
                except:
                    continue
            charactersTogether.append(prevChars)
            ind += 1
    else:
        print ""
    # do a quick check for multiples of the same character
    csToRem = []
    for c in charactersTogether[ind]:
        if charactersTogether[ind].count(c) > 1:
            csToRem.append(c)
    csToRem = list(set(csToRem))
    #print csToRem
    for c in csToRem:
        charactersTogether[ind].remove(c)
    #print "\t",charactersTogether[ind]
    
# with the characters listed, make the adjacency matrix
nChars = len(charNames)
adj = np.zeros((nChars,nChars))
# loop over all pairs of characters in each stage group
# make their connection in the adjancency matrix 1
# there will be tons of overlap here
for sitch in charactersTogether:
    for a,b in itertools.product(sitch,sitch):
        if a == b:
            continue
        inda = charNames.index(a)
        indb = charNames.index(b)
        adj[inda,indb] = 1
# make the graph and save it as a .gexf for Gephi
G = nx.Graph(adj)
G = nx.relabel_nodes(G,{i:n for i,n in enumerate(charNames)})
nx.write_gexf(G,"%s.gexf"%filename[:-4])