def gen_candidate(word):
"""
Generate candidate using k-gram index or automata
"""
global matcher, use_automata, bindex, word_dict, bindex_thresh, lang, cand_topk_cur
if use_automata:
cands = list(automata.find_all_matches(word, 2, matcher))
else:
word_aug = '$' + word + '$'
idx = range(len(word_aug))
query = set()
for tup in itertools.izip(idx[:-1],idx[1:]):
query.add(word_aug[tup[0]]+word_aug[tup[1]])
candidate = {}
query = [u for u in query if u in bindex]
for item in query:
lst = [word_dict[u] for u in bindex[item] if abs(len(word_dict[u])-len(word)) <= 2]
for w in lst:
if w in candidate:
candidate[w] += 1
else:
candidate[w] = 1
cands = []
for key, cnt in candidate.iteritems():
score = float(cnt) / float(len(key) + 1 + len(query) - cnt)
if score >= bindex_thresh and edit_distance(key, word) <= 2:
cands.append(key)
if len(cands) == 0 and word in lang:
cands.append(word)
cands = [(cand, edit_distance_plus(cand, word)[1]) for cand in cands]
cands = sorted(cands, key=lambda x: lang[x[0]] + x[1])
cands = [cand for cand, score in cands[:cand_topk_cur]]
return cands
def warui(lang, term, k=1):
"""Find bad words that matches input
Args:
lang: Language code such as en, es, fr, jp
term: Input word
k: Maximum edit distance
Yields:
Number of matches, matched words and probes as Tuple
"""
if not isinstance(term, unicode):
term = term.decode('utf-8')
words = get_words(lang)
m = Matcher(words)
li = list(automata.find_all_matches(term, k, m))
return (len(li), li, m.probes)
def candidates_for(self, obs, N = 10): if self.enable_dict and obs in self.dm.proper_noun_list: return [obs] if obs in self.d: # don't need to find candidates if valid word already return [obs] elif obs in self.memoize_can.keys(): return self.memoize_can[obs] else: candidates = list(automata.find_all_matches(obs, self.k, self.m)) if len(candidates) == 0: self.memoize_can[obs] = [obs] else: can_prob = [(c, self.emission_prob(obs, c)) for c in candidates] if self.enable_dict: can_prob.append((obs, self.emission_prob(obs, obs))) can_prob.sort(key = lambda x: x[1], reverse = True) # rank candidates by higher probability self.memoize_can[obs] = [can_prob[i][0] for i in range(min(N, len(can_prob)))] return self.memoize_can[obs]
def test_slice(self):
m = Matcher(words)
self.assertEqual(len((list(automata.find_all_matches('slice', 1, m)))), 15)
self.assertEqual(len((list(automata.find_all_matches('slice', 2, m)))), 213)
def test_hello(self):
m = Matcher(words)
self.assertEqual(len((list(automata.find_all_matches('hello', 1, m)))), 12)
self.assertEqual(len((list(automata.find_all_matches('hello', 2, m)))), 128)
def test_food(self):
m = Matcher(words)
self.assertEqual(len((list(automata.find_all_matches('food', 1, m)))), 18)
self.assertEqual(len((list(automata.find_all_matches('food', 2, m)))), 318)
import time
start_time = time.time()
words = [x.strip().lower().decode('utf-8').split() for x in open('../text/3.txt')]
words = [item for sublist in words for item in sublist]
words.sort()
# words10 = [x for x in words if random.random() <= 0.1]
# words100 = [x for x in words if random.random() <= 0.01]
# m = Matcher(words)
# print m.probes
m = Matcher(words)
print list(automata.find_all_matches('test', 2, m))
print m.probes
print("--- %s seconds ---" % (time.time() - start_time))
# def levenshtein(s1, s2):
# if len(s1) < len(s2):
# return levenshtein(s2, s1)
# if not s1:
# return len(s2)
# previous_row = xrange(len(s2) + 1)
# for i, c1 in enumerate(s1):
# current_row = [i + 1]
# for j, c2 in enumerate(s2):
self.probes = 0
def __call__(self, w):
self.probes += 1
pos = bisect.bisect_left(self.l, w)
if pos < len(self.l):
return self.l[pos]
else:
return None
f = open('known.json')
known = json.loads(f.read())
words = [x.strip().lower() for x in open('wordsEn.txt')]
m = Matcher(words)
results = []
for k, v in known.iteritems():
results.append(list(automata.find_all_matches(k, v, m)))
numresults = len(results)
common_set = set(results[0])
for i in range(numresults):
common_set.intersection_update(set(results[i]))
print sorted(common_set)
print len(common_set)
def __call__(self, w):
self.probes += 1
pos = bisect.bisect_left(self.l, w)
if pos < len(self.l):
return self.l[pos]
else:
return None
words = [x.strip().lower().decode('utf-8') for x in open('/usr/share/dict/web2')]
words.sort()
# words10 = [x for x in words if random.random() <= 0.1]
# words100 = [x for x in words if random.random() <= 0.01]
m = Matcher(words)
matches = list(automata.find_all_matches('food', 1, m))
assert len(matches) == 18
print m.probes
m = Matcher(words)
matches = list(automata.find_all_matches('food', 2, m))
assert len(matches) == 288
print m.probes
def levenshtein(s1, s2):
if len(s1) < len(s2):
return levenshtein(s2, s1)
if not s1:
return len(s2)
pos = bisect.bisect_left(self.l, w)
if pos < len(self.l):
return self.l[pos]
else:
return None
words = [x.strip().lower().decode('utf-8')
for x in open('/usr/share/dict/words')]
words.sort()
words10 = [x for x in words if random.random() <= 0.1]
words100 = [x for x in words if random.random() <= 0.01]
m = Matcher(words)
res = list(automata.find_all_matches('startinq', 1, m))
print res
print len(res)
print m.probes
"""
m = Matcher(words)
res = list(automata.find_all_matches('food', 2, m))
print res
print len(res)
print m.probes
"""
def levenshtein(s1, s2):
if len(s1) < len(s2):
else:
return None
words = [
x.strip().lower().decode('utf-8') for x in open('/usr/share/dict/words')
]
print 'l', len(words)
#words = [x.strip().lower().decode('utf-8') for x in open('/usr/share/dict/web2')]
words.sort()
words10 = [x for x in words if random.random() <= 0.1]
words100 = [x for x in words if random.random() <= 0.01]
m = Matcher(words)
from time import time
print len(list(automata.find_all_matches('food', 1, m)))
st = time()
print(list(automata.find_all_matches('mosco', 2, m)))
print 'st', time() - st
print m.probes
assert len(list(automata.find_all_matches('food', 1, m))) == 18
print m.probes
m = Matcher(words)
assert len(list(automata.find_all_matches('food', 2, m))) == 283
print m.probes
def levenshtein(s1, s2):
if len(s1) < len(s2):
return levenshtein(s2, s1)
start_time = time.time()
words = [
x.strip().lower().decode('utf-8').split() for x in open('../text/3.txt')
]
words = [item for sublist in words for item in sublist]
words.sort()
# words10 = [x for x in words if random.random() <= 0.1]
# words100 = [x for x in words if random.random() <= 0.01]
# m = Matcher(words)
# print m.probes
m = Matcher(words)
print list(automata.find_all_matches('test', 2, m))
print m.probes
print("--- %s seconds ---" % (time.time() - start_time))
# def levenshtein(s1, s2):
# if len(s1) < len(s2):
# return levenshtein(s2, s1)
# if not s1:
# return len(s2)
# previous_row = xrange(len(s2) + 1)
# for i, c1 in enumerate(s1):
# current_row = [i + 1]
# for j, c2 in enumerate(s2):
# insertions = previous_row[j + 1] + 1 # j+1 instead of j since previous_row and current_row are one character longer
def __call__(self, w):
self.probes += 1
pos = bisect.bisect_left(self.l, w)
if pos < len(self.l):
return self.l[pos]
else:
return None
words = [x.strip().lower() for x in open('test.txt')]
words.sort()
words10 = [x for x in words if random.random() <= 0.1]
words100 = [x for x in words if random.random() <= 0.01]
m = Matcher(words)
assert len(list(automata.find_all_matches('food', 1, m))) == 2
print(m.probes)
# m.probes
m = Matcher(words)
assert len(list(automata.find_all_matches('food', 2, m))) == 3
# print
# m.probes
print(list(automata.find_all_matches('food', 4, m)))
def levenshtein(s1, s2):
if len(s1) < len(s2):
return levenshtein(s2, s1)
if not s1: