matching two addresses via ngram one hot encoded cosine similarities

ngram_cosine.py

from nltk.util import ngrams
from sklearn.metrics.pairwise import cosine_similarity
import string
import itertools

vector_of_possibilities = [''.join(i) for i in itertools.product(string.ascii_lowercase + string.digits, repeat=3)]

def get_3grams(astring):
  newstring = [achar for achar in astring.lower() if achar.isalnum()]
  return [''.join(agram) for agram in ngrams(newstring, 3)]

def string_to_vec(astring):
  vec = [0] * len(vector_of_possibilities)
  for agram in get_3grams(astring):
      for i, apossibility in enumerate(vector_of_possibilities):
          if agram == apossibility:
              vec[i] = 1
  return [vec]

print(cosine_similarity(string_to_vec('48 Leabrooks Road Alfreton'),
                        string_to_vec(u'48 Leabrooks RdSomercotesAlfretonDE55 4HB')))