bag of words tanimoto.py

import re

NON_WORD = re.compile(r'\W+')

target = '55 w. 45th st.'
candidates = [
'11 park pl | btwn broadway & church st ',
'217 e 86th st | at 3rd ave ',
'2 e 33rd st | at 5th ave ',
"23 3rd ave | btwn st mark's pl & stuyvesant st",
'55 w 45th st | at 6th ave ',
'2047 broadway | btwn 70th & 71st st ',
'401 7th ave | btwn 32nd & 33rd st ',
'1 e 43rd st | btwn 5th & madison ave',
'122 university pl | btwn 13th & 14th st ']

def bag(sentence):
    words = sanitize(sentence).split()
    return set(words)

def tanimoto(a, b):
    return len(a.intersection(b)) * 1.0/ len(a.union(b))

def sanitize(s):
    return NON_WORD.sub(' ', s)

target_bag = bag(target)
for candidate in candidates:
    score = tanimoto(target_bag, bag(candidate))
    print('%f\t%s ~ %s' % (score, target, candidate))