Tries to ignore headings and template/variable/category tags, tries to count alternate link texts. Ignores a "Main Page" and "Template:" and "Category:" template types. This script is only reasonable if the dump is "current" page text only.

naive_wikidump_word_counts.py

import sys
import re
from pprint import pprint
from operator import itemgetter
from xml.dom import minidom

DEBUG = len(sys.argv[1:])

xml_filename = 'dump.xml'
ignore_titles = ['Main Page']
ignore_prefixes = ['Template', 'Category']

re_ignore_titles = re.compile(r'^((%s)|(%s):.+)$' % (
    '|'.join(ignore_titles),
    '|'.join(ignore_prefixes),
))
re_link = re.compile(r'\[\[(?!Category:)(.*?)\]\]')  # real links not category tagging
re_ignore = re.compile(r'(=+)\s*.*?\s*(\1)|(\{\{[^\}]+\}\})|(\[\[[^\]]+\]\])')  # headings/tags

def count_words(s):
    return len(s.split())

def count(data):
    n = 0
    for link_match in re_link.finditer(data):
        link_text = link_match.group(1).split('|')[-1]
        n += count_words(link_text)
        if DEBUG:
            print 'links:', n, ':', repr(link_text)
    stripped_data = re_ignore.sub('', data)
    n += count_words(stripped_data)
    if DEBUG:
        print 'all:', n, ':', repr(stripped_data.split())
    return n

d = minidom.parse(xml_filename)
counts = []
for page in d.getElementsByTagName('page'):
    if DEBUG:
        print '-------------'
    title = page.getElementsByTagName('title')[0].childNodes[0].data
    if re_ignore_titles.match(title):
        if DEBUG:
            print '(skipping %r)' % title
        continue
    if DEBUG:
        print title

    textNode = page.getElementsByTagName('text')[0]
    text = ' '.join(map(lambda x: x.data, textNode.childNodes))
    counts.append([title, count(text)])

total = sum(dict(counts).values())
number_size = len(str(total))
counts = sorted(counts, key=itemgetter(1), reverse=True)
for title, count in counts:
    print '%s: %s' % (str(count).rjust(number_size), title)
print '%s= ===========' % ('=' * number_size)
print '%s: Total words' % (str(total).rjust(number_size),)

~ python2 countwords.py
876: Big page
 10: Smaller page
  0: Page with a heading only
==== ===========
886: Total words