from lxml import etree
import re, codecs, nltk

f1 = codecs.open('media1.xml', 'r', 'utf-8')
doc = etree.parse('disamb-1.xml')
txt1 = f1.read().replace(u'=', u',').replace(u'S-PRO', u'SPRO').replace(u'A-PRO', u'APRO').replace('ADV-PRO', 'ADVPRO')
m = re.findall(u'gr,"([A-Za-z0-9,.-]*)">', txt1, flags = re.U)
for i in m:
    n = i.split(u',')
    s = sorted(n)
    gr = u','.join(s)
bg1 = nltk.bigrams(gr)
fd = nltk.FreqDist(bg1)


spisok = doc.findall('.//w')
arr = []
for i in spisok:
    k = i.getchildren()
    l = [child.get('gr') for child in k]
    arr.append(l)
n = 0
while n + 1 < len(arr):
    previous = arr[n][0]
    current = arr[n + 1]
    freqd = [(previous, i) for i in current]
    chisla = [fd[r] for r in freqd]
    idx = chisla.index(max(chisla))
    arr[n + 1] = [arr[n + 1][idx]]
    n += 1
print arr


f1.close()