def find_searchedperson(first_name, last_name, national_id):
'''
Try super hard to match this person up. Rather return bad
match than no match - we can manually search bad matches.
'''
if national_id in persons_by_id:
yield persons_by_id[national_id]
last_name_norm = normalize_string(last_name)
matches = []
if last_name_norm in persons_by_last_name:
matches = persons_by_last_name[last_name_norm]
else:
for key in persons_by_last_name.keys():
if jaro(last_name_norm, key) > 0.9:
matches.extend(persons_by_last_name[key])
first_name_norm = normalize_string(first_name)
for match in matches:
key = normalize_string(match.first_name)
'''
A match is valid if:
1. We don't have a first name because only last name
was used in a failed search.
2. The entire first name string is similar.
3. One of the first names are similar.
'''
if (not key and isinstance(match, SearchedPersonNotFound)) \
or jaro(first_name_norm, key) > 0.9 \
or has_matching_word(first_name_norm, key):
yield match
def compare_two_names(name1, name2, max_splits=7):
def normalize_name(s):
return re.sub(r"\s+", " ", s.lower().strip().replace("-", " "))
def slugify_name(s):
return (
s.replace(" ", "")
.replace(".", "")
.replace('"', "")
.replace("'", "")
.replace("’", "")
)
name1 = normalize_name(name1)
name2 = normalize_name(name2)
if slugify_name(name1) == slugify_name(name2):
return 1
if jaro(name1, name2) > 0.95:
return 1
splits = name2.split(" ")
limit = reduce(mul, range(1, max_splits + 1))
if len(splits) > max_splits:
print("Too much permutations for {}".format(name2))
return max(
jaro(name1, " ".join(opt)) for opt in islice(permutations(splits), limit)
)
def full_compare(name1, name2):
name1 = _normalize_name(name1)
name2 = _normalize_name(name2)
slugified_name1 = _slugify_name(name1)
slugified_name2 = _slugify_name(name2)
if slugified_name1 == slugified_name2:
return True
if slugified_name1.startswith(slugified_name2) and len(slugified_name2) >= 10:
return True
if slugified_name2.startswith(slugified_name1) and len(slugified_name1) >= 10:
return True
if slugified_name1.endswith(slugified_name2) and len(slugified_name2) >= 10:
return True
if slugified_name2.endswith(slugified_name1) and len(slugified_name1) >= 10:
return True
if jaro(slugified_name1, slugified_name2) > 0.95:
return True
if jaro(slugified_name2, slugified_name1) > 0.95:
return True
if _compare_two_names(name1, name2):
return True
if _compare_two_names(name2, name1):
return True
return _thorough_compare(name1, name2) or _thorough_compare(name2, name1)
def find_searchedperson(first_name, last_name, national_id):
'''
Try super hard to match this person up. Rather return bad
match than no match - we can manually search bad matches.
'''
if national_id in persons_by_id:
yield persons_by_id[national_id]
last_name_norm = normalize_string(last_name)
matches = []
if last_name_norm in persons_by_last_name:
matches = persons_by_last_name[last_name_norm]
else:
for key in persons_by_last_name.keys():
if jaro(last_name_norm, key) > 0.9:
matches.extend(persons_by_last_name[key])
first_name_norm = normalize_string(first_name)
for match in matches:
key = normalize_string(match.first_name)
'''
A match is valid if:
1. We don't have a first name because only last name
was used in a failed search.
2. The entire first name string is similar.
3. One of the first names are similar.
'''
if (not key and isinstance(match, SearchedPersonNotFound)) \
or jaro(first_name_norm, key) > 0.9 \
or has_matching_word(first_name_norm, key):
yield match
def get_jaro_to_list(first4jaro, list4jaro, factor=0.9):
result = [[0 for x in range(len(list4jaro))]
for y in range(len(first4jaro))]
loc_data = 0.0
#If loc_data =0, we take the first one
loc_i = 0
loc_j = 0
for i, item in enumerate(first4jaro):
for j, data in enumerate(list4jaro):
if (item[1] == "") or (data[1] == ""):
result[i][j] = jaro(item[0], data[0])
else:
result[i][j] = jaro(item[0], data[0]) * jaro(item[1], data[1])
if result[i][j] > loc_data:
loc_data = result[i][j]
loc_i = i
loc_j = j
first2return = first4jaro[:loc_i] + first4jaro[loc_i + 1:]
list4return = list4jaro[:loc_j] + list4jaro[loc_j + 1:]
if (len(first2return) == 0) or (len(list4return) == 0):
dif = abs(len(first2return) - len(list4return))
return loc_data * loc_data * math.pow(factor, dif)
else:
return loc_data * loc_data * get_jaro_to_list(first2return,
list4return)
def _get_name(name, first_name, gender):
for prefix in ('eng', 'dr', 'hr', 'phd'):
if first_name.startswith(prefix + ' '):
first_name = first_name[len(prefix) + 1:]
first_name = first_name.replace(' ', '')
if first_name in ('abd', 'abdel', 'عبد') or first_name == '':
ns = name.split(' ')
segs = []
for s in ns:
segs.append(s)
if s not in ('abd', 'el', 'عبد'):
break
first_name = ' '.join(segs).replace(' ', '')
if is_arabic(first_name):
return first_name
if 'female' in gender:
gender = 'female'
elif 'male' in gender:
gender = 'male'
if gender == 'unknown':
if first_name in males_en or first_name in males:
gender = 'male'
elif first_name in females_en or first_name in females:
gender = 'female'
if gender == 'unknown':
nearest = -1
nearest_gender = 'unknown'
for en_name in males_en:
similarity = jaro(first_name, en_name)
if similarity > nearest:
nearest_gender = 'male'
nearest = similarity
if similarity == nearest and nearest_gender == 'female':
nearest_gender = 'unknown'
for en_name in females_en:
similarity = jaro(first_name, en_name)
if similarity > nearest:
nearest_gender = 'female'
nearest = similarity
if similarity == nearest and nearest_gender == 'male':
nearest_gender = 'unknown'
gender = nearest_gender
d = males if gender == 'male' else females if gender == 'female' else unknowns
if first_name in d:
return d[first_name]
res = translator.translate(first_name, dest='ar', src='en').text
d[first_name] = res
return res
def get_compared_data_file(data, language="en", data_kind="surname"):
'''
This function will compare the given name with the current data input
'''
if language in LANGUAGES_FILES.keys():
if data_kind in LANGUAGES_FILES[language].keys():
data_in_met = adapted_doublemetaphone(data, language=language)
total_data = []
for word, met_value in LANGUAGES_DATA[language][data_kind].items():
if met_value == data_in_met:
total_data.append(word)
#If the value is already available, we just return it
if data in LANGUAGES_DATA[language][data_kind].keys():
return data, 1.0
else:
data_temp = data.lower()
norm = LANGUAGES_FILES[language]["normalize"]
for notnorm in norm.keys():
data_temp = data_temp.replace(notnorm, norm[notnorm])
results = {}
for candidate in total_data:
candidate_temp = candidate.lower()
for notnorm in norm.keys():
candidate_temp = candidate_temp.replace(
notnorm, norm[notnorm])
results[candidate] = jaro(candidate_temp, data_temp)
if (any(results)):
return max(results, key=results.get), max(results.values())
else:
return data, -1.0
return data, -1.0
def score_reconciliation(txn, payment):
words = list(filter(None, re.split(r"\W+", txn.payee)))
bankref_parts = [payment.bankref[:4], payment.bankref[4:]]
bankref_distances = [ratio(w, p) for w in words for p in bankref_parts]
# Get the two best matches, for the two parts of the bankref
# A match gives 1.0, a 2-char substring 0.666, and a 6-char superstring 0.857
bankref_score = sum(sorted(bankref_distances)[-2:])
name_score = jaro(txn.payee, payment.user.name)
other_score = 0.0
if txn.amount == payment.amount:
other_score += 0.4
if txn.account.currency == payment.currency:
other_score += 0.6
# check posted against expiry?
app.logger.debug(
"Scores for txn %s payment %s: %s %s %s",
txn.id,
payment.id,
bankref_score,
name_score,
other_score,
)
return bankref_score + name_score + other_score
def update(self, idx, idy, get_word):
# ox = self.store.get(idx)
# oy = self.store.get(idy)
#save distance:
# self.vec[idx,idy] =
stime = datetime.now()
try:
idx = int(idx)
idy = int(idy)
if idx == idy:
return
ox = self.store.get(str(idx))
oy = self.store.get(str(idy))
if ox is None or oy is None:
return
w1 = get_word(ox)
w2 = get_word(oy)
simi = jaro(w1, w2)
logger.debug("({},{}) vs ({},{}) : {}".format(
idx, w1, idy, w2, simi))
self.store.set_entry(idx, idy, simi)
except Exception as e:
logger.debug("(idx : {} type : {})".format(idx, type(idx)))
raise e
ntime = datetime.now()
logger.debug("processing time : {}".format(ntime - stime))
def full_compare(name1, name2):
def normalize_name(s):
return re.sub(r"\s+", " ", s.strip().replace("-", " "))
def slugify_name(s):
s = (s.replace(" ", "").replace(".", "").replace('"', "").replace(
"'",
"").replace("’", "").replace("є", "е").replace("i", "и").replace(
"ь", "").replace("'",
"").replace('"', "").replace('`', "").replace(
"’", "").replace("ʼ", ""))
return re.sub(r"\d+", "", s)
name1 = normalize_name(name1)
name2 = normalize_name(name2)
slugified_name1 = slugify_name(name1)
slugified_name2 = slugify_name(name2)
if slugified_name1 == slugified_name2:
return True
if slugified_name1.startswith(
slugified_name2) and len(slugified_name2) >= 10:
return True
if slugified_name2.startswith(
slugified_name1) and len(slugified_name1) >= 10:
return True
if jaro(slugified_name1, slugified_name2) < 0.6:
return False
if jaro(slugified_name1, slugified_name2) > 0.95:
return True
if jaro(slugified_name2, slugified_name1) > 0.95:
return True
if _compare_two_names(name1, name2):
return True
if _compare_two_names(name2, name1):
return True
return False
def fetch_answer(Q_input):
score_list = list(map(lambda x: jaro(x, Q_input), Q_list))
highest_score = max(score_list)
highest_score_index = score_list.index(highest_score)
selected_answer = A_list[highest_score_index]
score_list.pop(highest_score_index)
second_highest_score = max(score_list)
return selected_answer, highest_score, second_highest_score
def predict_ner():
'''
This will return the similarity between two strings
'''
str1 = str(request.form['string1'])
str2 = str(request.form['string2'])
result = jaro(str1, str2)
return render_template('ner.html', prediction_text='{}'.format(result))
def cmp_auth(self, s1, s2):
if len(s1) != len(s2): return False, False
matching, sorted = True, True
sim_mat = [[jaro(s1[y], s2[x]) for x in range(len(s2))]
for y in range(len(s1))]
for i in range(len(s2)):
if max(sim_mat[i]) != sim_mat[i][i]: sorted = False
if max(sim_mat[i]) < self.threshold['author']: matching = False
return matching, sorted
def compare_fingerprints(left, right):
result = 0
left_list = ensure_list(left.get('fingerprints'))
right_list = ensure_list(right.get('fingerprints'))
for (left, right) in itertools.product(left_list, right_list):
similarity = jaro(left, right)
score = similarity * dampen(3, 20, min(left, right, key=len))
result = max(result, score)
return result
def compare_names(left, right):
result = 0
left_list = list(_normalize_names(left.names))
right_list = list(_normalize_names(right.names))
for (left, right) in itertools.product(left_list, right_list):
similarity = jaro(left, right)
score = similarity * dampen(2, 20, shortest(left, right))
result = max(result, score)
return result
def compare_names(left, right):
result = 0
left_list = [normalize(n, latinize=True) for n in left.names]
right_list = [normalize(n, latinize=True) for n in right.names]
for (left, right) in itertools.product(left_list, right_list):
similarity = jaro(left, right)
score = similarity * dampen(2, 20, shortest(left, right))
result = max(result, score)
return result
def get_classify(tuple_cont):
'''判断当前站点所属类别,该函数为进程池目标函数
'''
try:
if jaro(Config.ipc_r_kw.value[0], tuple_cont[0]) > Config.edit_dist_benchmark:
Config.ipc_list_url.append(tuple_cont[1] + '\n')
except:
# print '#############################################'
# print Config.ipc_r_kw.value[1]
pass
def get_name_from_fullname(full_name,
list_father_surnames,
list_mother_surnames,
language="en"):
'''
Given a full name, including surname, this function will provide out the first name of
the person removing the surname of the person
'''
merged_list = list_father_surnames + list_mother_surnames
for surname in merged_list:
temp_surname = surname.split(" ")
if len(temp_surname) > 1:
for i, _ in enumerate(temp_surname):
if temp_surname[i] in LANGUAGES_ADDS[language]:
temp_surname[i] = ""
new_surname = " ".join(temp_surname).rstrip().strip()
if (not new_surname in merged_list):
merged_list.append(new_surname)
merged_metaphore = []
for data in merged_list:
if adapted_doublemetaphone(data, language) not in merged_metaphore:
merged_metaphore.append(adapted_doublemetaphone(data, language))
full_name_list = get_splitted_name_from_complete_name(full_name, language)
for i, value in enumerate(full_name_list[0]):
#We remove from the specific particle the particles from each language that are used inside surnames
#to connect
check_surname = value.split(" ")
if len(check_surname) > 1:
for j, value in enumerate(check_surname):
if (check_surname[j].lower() in LANGUAGES_ADDS[language]):
check_surname[j] = ""
adapted_surname = "".join(check_surname).rstrip()
if (adapted_doublemetaphone(value, language)
in merged_metaphore) or (adapted_doublemetaphone(
adapted_surname, language) in merged_metaphore):
#The methapone algorithm is not perfect... so that we include here a crosschecking of very close phonetical, but far written data.
similar = 0
for compared in merged_list:
if jaro(adapted_surname, compared) > similar:
similar = jaro(adapted_surname, compared)
if similar > THRESHOLD_JARO:
full_name_list[0][i] = ""
return " ".join(full_name_list[0]).rstrip()
def fuzzyStr(perc, str_one, str_two):
if all(i in str_two.split() for i in str_one.split()):
return True
try:
perc = int(perc)
except:
return False
if int(jaro(str_one, str_two) * 100) >= perc:
return True
else:
return False
def score_of_given_name_and_meta(first4jaro,
list4jaro,
name1,
name2,
factor=0.9):
'''
This function will take the maximum score between the direct comparison of the name and the phonetic comparison
'''
score_compare = jaro(name1, name2)
score_met = get_jaro_to_list(first4jaro, list4jaro, factor=factor)
return max(score_met, score_compare * score_compare)
def compare(x, y):
if x is None:
return [0, x, y]
x = normalize(x)
x_tokens = [word.lower() for word in tokenize(x) if word.isalnum()]
y = normalize(y)
y_tokens = [word.lower() for word in tokenize(y) if word.isalnum()]
ppx = untokenize(x_tokens)
ppy = untokenize(y_tokens)
return [jaro(ppx, ppy), ppx, ppy]
def _compare_two_names(name1,
name2,
max_splits=7,
straight_limit=0.93,
smart_limit=0.95):
splits = name2.split(" ")
straight_similarity = jaro(name1, name2)
if straight_similarity > smart_limit:
return True
if straight_similarity > straight_limit:
min_pair_distance = 1
for a, b in zip_longest(name1.split(" "), splits):
if a is not None and b is not None:
min_pair_distance = min(jaro(a, b), min_pair_distance)
if min_pair_distance > 0.8:
if len(splits) > 1 and DEBUG:
tqdm.write("Hmmm, looks like a match {}\t{}".format(
name1, name2))
return True
else:
if len(splits) > 1 and DEBUG:
tqdm.write("Check if it's match: {}\t{}".format(name1, name2))
limit = reduce(mul, range(1, max_splits + 1))
if len(splits) > max_splits and DEBUG:
tqdm.write("Too much permutations for {}".format(name2))
max_similarity = max(
jaro(name1, " ".join(opt))
for opt in islice(permutations(splits), limit))
return max_similarity > smart_limit
def score_of_given_name_and_meta(first4jaro,
list4jaro,
name1,
name2,
factor=0.9):
'''
This function will take the maximum score between the direct comparison of the name and the phonetic comparison
'''
#Jaro is creating odd situations with names which are very different in length, with this modification, we penalize lenght differences a lot
len_factor = (abs((len(name1) - len(name2))) / max(len(name1), len(name2)))
score_compare = jaro(name1, name2)
score_met = get_jaro_to_list(first4jaro, list4jaro, factor=factor)
if (len_factor < 0.33) or (1 - len_factor) * (1 - len_factor) > max(
score_met, score_compare * score_compare):
return max(score_met, score_compare * score_compare)
#We undo only in case this new scoring is more negative
else:
return (1 - len_factor) * (1 - len_factor)
def test_parse(self):
for file in os.listdir(SAMPLE_DIR):
if not file.endswith(".rst"):
continue
filename = os.path.join(SAMPLE_DIR, file)
article = parse_article(filename)
rendered = article.render().strip()
with open(filename) as f:
source = f.read().strip()
source = source.expandtabs(4).decode("utf8")
if source != rendered:
lev_ = distance(source, rendered)
jaro_ = jaro(source, rendered)
if lev_ > 10 and jaro_ < 0.8 and file not in MUTATED_FILES:
print("%d %f %s" % (lev_, jaro_, filename))
raise AssertionError(filename)
def _compare_two_names(
name1, name2, max_splits=7, straight_limit=0.70, smart_limit=0.96
):
straight_similarity = jaro(name1, name2)
if straight_similarity > smart_limit:
return True
if straight_similarity > straight_limit:
min_pair_distance = 1
for a, b in zip_longest(name1.split(" "), name2.split(" ")):
if a is not None and b is not None:
chunk_distance = _smart_jaro(a, b, func=jaro_winkler)
min_pair_distance = min(chunk_distance, min_pair_distance)
if min_pair_distance > 0.88:
return True
return False
def match_location(self, location):
"""
We will mutate the score a bit to add +0.1 to the jaro distance
for starting with the same letter. Alexa's speech processing system
really sucks at this.
location -- the location to match against
"""
if location:
matches = []
for switch_id, switch_func in self.server.switches.items():
similarity = jaro(location, switch_id)
if location[0].lower() == switch_id[0].lower():
similarity += 0.1
matches += [(similarity, switch_id, switch_func)]
matches.sort(key=lambda x: x[0], reverse=True)
if matches[0][0] >= _MATCH_THRESHOLD:
return matches[0][1:]
raise ActionParseError("I didn't understand the location. "
"Could you please repeat?")
def annotate(self, training_set):
#Levenshtein distance - minimum number of single character edits
distance_udf = udf(lambda x, y: distance(x, y), IntegerType())
#Levenshtein ratio - similarity of two strings
ratio_udf = udf(lambda x, y: ratio(x, y), DoubleType())
#Jaro - similarity score
jaro_udf = udf(lambda x, y: jaro(x, y), DoubleType())
#Jaro-winkler - similarity score, which favors strings that match prefix from the beginning
jaro_winkler_udf = udf(lambda x, y: jaro_winkler(x, y), DoubleType())
#fuzz partial ratio - gives a score based on how well parts of a string match another
fuzz_partial_ratio_udf = udf(
lambda x, y: fuzz.partial_ratio(x, y) / 100, DoubleType())
training_set = training_set.withColumn("distance", distance_udf("concept_name_1", "concept_name_2")) \
.withColumn("ratio", ratio_udf("concept_name_1", "concept_name_2")) \
.withColumn("jaro", jaro_udf("concept_name_1", "concept_name_2")) \
.withColumn("jaro_wrinkler", jaro_winkler_udf("concept_name_1", "concept_name_2")) \
.withColumn("fuzz_partial_ratio", fuzz_partial_ratio_udf("concept_name_1", "concept_name_2"))
return training_set
def score_reconciliation(txn, payment):
words = txn.payee.replace('-', ' ').split(' ')
bankref_distances = [ratio(w, payment.bankref) for w in words]
# Get the two best matches, for the two parts of the bankref
bankref_score = sum(sorted(bankref_distances)[-2:])
name_score = jaro(txn.payee, payment.user.name)
other_score = 0.0
if txn.amount == payment.amount:
other_score += 0.4
if txn.account.currency == payment.currency:
other_score += 0.6
# check posted against expiry?
app.logger.debug('Scores for txn %s payment %s: %s %s %s',
txn.id, payment.id, bankref_score, name_score, other_score)
return bankref_score + name_score + other_score
def match_locality(string, localities):
''' Try to figure out which locality 'string' is by
finding the known localities that have the maximum
(jaro) similarity score '''
if ',' in string:
parts = string.split(',')
string = parts[1].strip() + ' ' + parts[0]
best = 0
matches = []
jaro_hits = []
substring_hits = []
for loc in localities:
ulower_string = strip_accents(string.lower().decode("utf-8"))
ulower_loc_name = strip_accents(loc['loc_name'].lower().decode("utf-8"))
smaller, bigger = ulower_string, ulower_loc_name
if len(bigger) < len(smaller):
smaller, bigger = bigger, smaller
if len(smaller) < len(bigger): # they might actually be the same size
if smaller in bigger:
hit = dict(loc)
hit['score'] = 'sub'
substring_hits.append(hit)
similarity = jaro(ulower_string, ulower_loc_name)
if similarity > best:
matches = []
best = similarity
if similarity == best:
hit = dict(loc)
hit['score'] = similarity
matches.append(hit)
jaro_hits = [(m['loc_name'],m['muni_name']) for m in matches]
for s in substring_hits:
if (s['loc_name'],s['muni_name']) not in jaro_hits:
matches.append(s)
return matches
def score_reconciliation(txn, payment):
words = list(filter(None, re.split('\W+', txn.payee)))
bankref_parts = [payment.bankref[:4], payment.bankref[4:]]
bankref_distances = [ratio(w, p) for w in words for p in bankref_parts]
# Get the two best matches, for the two parts of the bankref
# A match gives 1.0, a 2-char substring 0.666, and a 6-char superstring 0.857
bankref_score = sum(sorted(bankref_distances)[-2:])
name_score = jaro(txn.payee, payment.user.name)
other_score = 0.0
if txn.amount == payment.amount:
other_score += 0.4
if txn.account.currency == payment.currency:
other_score += 0.6
# check posted against expiry?
app.logger.debug('Scores for txn %s payment %s: %s %s %s',
txn.id, payment.id, bankref_score, name_score, other_score)
return bankref_score + name_score + other_score
async def special_matches(self, monster):
max_score = 0
for class_attrs in self.monster_class_attributes:
val: MonsterModel = monster
for class_attr in class_attrs:
val: str = getattr(val, class_attr, None)
if val is None:
continue
val: str = val.lower()
if self.match == "=" and val == self.string: # Exact match
return True, MatchData(self)
elif self.match == "r" and bool(re.search(self.string,
val)): # Regex match
return True, MatchData(self)
elif self.match == "g" and fnmatch(
val, '*' + self.string + '*'): # Glob match
return True, MatchData(self)
elif self.string in val:
return True, MatchData(self)
max_score = max(max_score, jaro(self.string, val))
if max_score >= TOKEN_JW_DISTANCE:
return max_score, MatchData(self)
return False, MatchData(self)
def get_common_audios(login: str, password: str, *ids):
#пока только для 2 акков
#добавить lru_cache
vk_session = vk_api.VkApi(login, password, app_id=app_id, scope=scope)
vk_session.auth()
#vk=vk_session.get_api()
vkaudio = audio.VkAudio(vk_session)
users_track_list = []
for id in ids:
user_track_list = set([
track['artist'] + '-' + track['title']
for track in vkaudio.get_iter(id)
])
print("I'm not dead")
users_track_list.append(user_track_list)
shares = [len(i) for i in users_track_list]
common_audios_lst = set()
while (len(users_track_list) > 1):
common_audios_lst = set()
users_track_list.sort(key=lambda x: len(x))
for i in users_track_list[0]:
for j in users_track_list[1]:
if jaro(i, j) > 0.75:
common_audios_lst.add(i)
users_track_list.pop(0)
print("I'm not dead")
users_track_list[0] = common_audios_lst
print("_______________________")
print(common_audios_lst)
shares = [len(common_audios_lst) / i for i in shares.copy()]
returnable_text = ''
for i in common_audios_lst:
returnable_text += (i + '\n')
returnable_text += "Процент общих песен\n"
for i in shares:
returnable_text += (str(round(100 * i, 2)) + r'% ')
return returnable_text[:-1]
ddir = '/home/ngaude/workspace/data/cdiscount/'
test = pd.read_csv(ddir+'test.csv',sep=';').fillna('');
test['lib'] = map(normalize_guess,test.Libelle.values)
test = test.sort('lib').reset_index(drop=True)
resultat = pd.read_csv(ddir+'test.csv',sep=';').fillna('');
a = test.lib.values
b = [0]
for i in range(0,len(a)-1):
if len(a[i])<8 or len(a[i+1])<8:
b.append(0)
else:
b.append(jaro(a[i],a[i+1]))
"""
plt.hist(b,bins=300,cumulative=True)
plt.show()
"""
cut_threshold = np.percentile(b,50)
same_categorie_than_previous_item = [i>cut_threshold for i in b]
group_categorie = [0]*len(same_categorie_than_previous_item)
for i in range(1,len(same_categorie_than_previous_item)):
if same_categorie_than_previous_item[i] == True:
group_categorie[i] = group_categorie[i-1]
else:
word_phonemes = dict()
matched = dict()
# Map words to their phonemes
for i in range(len(words)):
word_phonemes[words[i]] = phonemes[i]
print ("word1,word2,semantic_similarity,phonetic_similarity,word_similarity,sem_x_phon_similarity")
for triade in semsim:
a = triade[0]
b = triade[1]
# ignore duplicated (b,a)
dup = matched.get(b + "_" + a, None)
if dup:
continue
matched[a + "_" + b] = True
# ignore pairs with the same stem
if (stem(a) == stem(b)):
continue
# Get their phonemes
ph1 = word_phonemes.get(a, None)
ph2 = word_phonemes.get(b, None)
if ph1 is None or ph2 is None:
continue
# Semantic similarity
ss = float(triade[2])
# Phonetic similarity
ps = jaro(ph1, ph2)
# Word similarity
ld = jaro(a,b)
print ("%s,%s,%.4f,%.4f,%.4f,%.4f" % (a,b, ss, ps, ld, ss*ps))
if office_holder_dict not in columbus_file:
columbus_file.append(office_holder_dict)
#puts detroit at large ***office holder*** dicts in a separate dictList
elif office_holder_dict['OCDID'] == 'ocd-division/country:us/state:mi/place:detroit' and office_holder_dict['Office Name'] != 'Mayor':
if office_holder_dict not in detroit_file:
detroit_file.append(office_holder_dict)
#puts boston at large ***office holder*** dicts in a separate dictList
elif office_holder_dict['OCDID'] == 'ocd-division/country:us/state:ma/place:boston' and office_holder_dict['Office Name'] != 'Mayor':
if office_holder_dict not in boston_file:
boston_file.append(office_holder_dict)
#if the dicts have UIDs, and are not charlotte at large, detroit at large, boston at large, or columbus council members, then start string comparison
else:
if scraped_dict['UID'] == office_holder_dict['UID']:
if scraped_dict['official.name'] == office_holder_dict['Official Name']:
print office_holder_dict['UID'],"scraped name: ", scraped_dict['official.name'], "file name: ",office_holder_dict['Official Name'], '\n\t>>>all good, exact match'
elif jaro(scraped_dict['official.name'].lower().replace(' ', '').replace('"','').replace('.','').replace(',',''),office_holder_dict['Official Name'].lower().replace(' ', '').replace('"','').replace('.','').replace(',','')) > .65:
print office_holder_dict['UID'], scraped_dict['official.name'],office_holder_dict['Official Name'], '\n\t>>>not exact match, but high lev score'
else:
print "\n\t>>>found a difference!"
print office_holder_dict['UID'],"scraped name: ", scraped_dict['official.name'], "file name: ",office_holder_dict['Official Name']
print jaro(scraped_dict['official.name'],office_holder_dict['Official Name'])
#answer = raw_input("\n\t>>>is this a meaningful difference? Y/N")
#if answer == "Y" or answer == "y":
checkList.append(office_holder_dict['UID'])
###output from initial scrape compare string comparisons
txt_file.append("\nCheck List: "+ ",".join(checkList))
txt_file.append("\nNo UID for:"+ ",".join(GPmissingList))
def handle(self, *args, **options):
activate(settings.LANGUAGE_CODE)
all_companies = []
keys = ["pk", "code", "name", "name_en", "short_name", "short_name_en"]
for p in Company.objects.all():
all_companies.append(dict(zip(keys, [
p.pk,
p.edrpou,
p.name_uk,
p.name_en,
p.short_name_uk,
p.short_name_en,
])))
grouped_by_code = defaultdict(list)
grouped_by_name = defaultdict(list)
# First pass: exact matches by code, full name or short name
for l in all_companies:
code = self.cleanup(l["code"])
if len(code) > 2:
grouped_by_code[code].append(l["pk"])
for k in ["name", "name_en", "short_name", "short_name_en"]:
name = self.cleanup(l[k])
if len(name) > 3:
grouped_by_name[name].append(l["pk"])
spoiled_ids = set()
chunks_to_review = list()
for k, v in grouped_by_code.items():
if len(set(v)) > 1:
spoiled_ids |= set(v)
chunks_to_review.append(v)
for k, v in grouped_by_name.items():
if len(set(v)) > 1:
spoiled_ids |= set(v)
chunks_to_review.append(v)
for chunk in chunks_to_review:
try:
CompanyDeduplication(
company1_id=chunk[0],
company2_id=chunk[1],
company1_json=Company.objects.get(pk=chunk[0]).to_dict(),
company2_json=Company.objects.get(pk=chunk[1]).to_dict(),
).save()
except IntegrityError:
pass
candidates_for_fuzzy = [
l for l in all_companies
if l["pk"] not in spoiled_ids
]
for a, b in combinations(candidates_for_fuzzy, 2):
for field_a, field_b in product(["name", "short_name"], repeat=2):
val_a = self.cleanup(a[field_a])
val_b = self.cleanup(b[field_b])
if len(val_a) < 4 or len(val_b) < 4:
continue
if self.cleanup_digits(a[field_a]) == self.cleanup_digits(b[field_b]):
continue
score = jaro(val_a, val_b)
if score > 0.97:
try:
CompanyDeduplication(
company1_id=a["pk"],
company2_id=b["pk"],
company1_json=Company.objects.get(pk=a["pk"]).to_dict(),
company2_json=Company.objects.get(pk=b["pk"]).to_dict(),
fuzzy=True,
).save()
break
except IntegrityError:
pass
for field_a, field_b in product(["name_en", "short_name_en"], repeat=2):
val_a = self.cleanup(a[field_a])
val_b = self.cleanup(b[field_b])
if len(val_a) < 4 or len(val_b) < 4:
continue
if self.cleanup_digits(a[field_a]) == self.cleanup_digits(b[field_b]):
continue
score = jaro(val_a, val_b)
if score > 0.97:
try:
CompanyDeduplication(
company1_id=a["pk"],
company2_id=b["pk"],
company1_json=Company.objects.get(pk=a["pk"]).to_dict(),
company2_json=Company.objects.get(pk=b["pk"]).to_dict(),
fuzzy=True,
).save()
break
except IntegrityError:
pass
def has_matching_word(phrase1, phrase2):
for word1 in phrase1.split():
for word2 in phrase2.split():
if jaro(word1, word2) > 0.9:
return True
def _jaro(a,b):
"""Jaro
The Jaro string similarity metric is intended for short strings like personal last names."""
return jaro(a,b)
def has_matching_word(phrase1, phrase2):
for word1 in phrase1.split():
for word2 in phrase2.split():
if jaro(word1, word2) > 0.9:
return True
def run(self, entry):
self.logger.create(entry['ID'])
article = search(entry)
crit, missing, support = self.check_tags(entry)
if not support:
self.logger.log(
'No support for ' + entry['ENTRYTYPE'] +
'. Currently, only the following entry types are supported: ' +
', '.join(set(self.tags.keys()).symmetric_difference({'all'})),
2)
self.logger.print()
return entry
if not crit:
self.logger.log('missing critical tag(s): ' + ' '.join(missing), 3)
self.logger.print()
return entry
if article is None:
self.logger.log('article could not be found on PubMed', 2)
self.logger.print()
return entry
if len(missing) > 0:
self.logger.log(
'the following tags are missing: ' + ' '.join(missing), 1)
self.logger.unindent()
# compare authors list
matching, sorted = self.cmp_auth(self.parse_auth(entry['author']),
article.authors)
if not matching:
self.logger.log('authors mismatch:', 3)
self.logger.log('PM: ' + ' & '.join(article.authors), 3)
self.logger.unindent()
self.logger.log(
'bib: ' + ' & '.join(self.parse_auth(entry['author'])), 3)
self.logger.unindent()
self.logger.unindent()
entry['author'] = self.pack_auth(article.authors)
elif not sorted:
self.logger.log('authors list misordered:', 3)
self.logger.log('PM: ' + ' & '.join(article.authors), 3)
self.logger.unindent()
self.logger.log(
'bib: ' + ' & '.join(self.parse_auth(entry['author'])), 3)
self.logger.unindent()
self.logger.unindent()
entry['author'] = self.pack_auth(article.authors)
# critical comparisons
for tag in set(self.tags[entry['ENTRYTYPE']]).symmetric_difference(
{'author'}):
idx = [entry[tag]]
if tag in self.aliases.keys():
for i in self.aliases[tag].keys():
if jaro(idx[0].strip(stripsym).lower(),
i.strip(stripsym).lower()) > self.threshold[tag]:
idx.append(self.aliases[tag][i])
if not any([
self.threshold[tag] <= jaro(
article.__dict__[dictionary[tag]].strip(
stripsym).lower(),
i.strip(stripsym).lower()) for i in idx
]):
self.logger.log(tag + ' mismatch:', 3)
self.logger.log(
'PM: ' +
article.__dict__[dictionary[tag]].strip(stripsym), 3)
self.logger.unindent()
self.logger.log('bib: ' + entry[tag].strip(stripsym), 3)
self.logger.unindent()
self.logger.unindent()
entry[tag] = article.__dict__[dictionary[tag]]
self.logger.print()
return entry
def discogs_ordered_search(query, item_type, limit=100):
name_pattern = ' \([0-9]+\)'
q_stripped = query.strip("'\"")
# special case when searching directly by id
if q_stripped.isdigit():
url = 'http://{host}/{item_type}s/{query}'.format(
host=DISCOGS_HOST,
query=urllib.quote_plus(query.lower()),
item_type=item_type
)
log.debug('search by id: {0}'.format(url))
r = requests.get(url)
if not r.status_code == 200:
return []
data = json.loads(r.text.replace('api.discogs.com', DISCOGS_HOST))
# TODO: not very nice - remap some fields
if item_type == 'release':
if 'title' in data:
data['title'] = re.sub(name_pattern, '', data['title'])
if 'formats' in data:
formats = []
for format in [f['name'] for f in data['formats'] if 'name' in f]:
formats.append(format)
data['format'] = formats
if 'labels' in data:
try:
data['catno'] = data['labels'][0]['catno']
except KeyError:
pass
if item_type == 'artist':
if 'name' in data:
data['title'] = re.sub(name_pattern, '', data['name'])
if 'aliases' in data:
aliases = []
for alias in [a['name'] for a in data['aliases'] if 'name' in a]:
aliases.append(re.sub(name_pattern, '', alias))
data['aliases'] = aliases
if 'members' in data:
members = []
for member in [m['name'] for m in data['members'] if 'name' in m]:
members.append(re.sub(name_pattern, '', member))
data['members'] = members
if 'images' in data:
for image in [i['uri150'] for i in data['images'] if 'type' in i and i['type'] == 'primary']:
data['thumb'] = image
break
return [data,]
url = 'http://{host}/database/search?q={query}&type={item_type}&per_page=100'.format(
host=DISCOGS_HOST,
query=urllib.quote_plus(query.encode('utf8').lower()),
item_type=item_type
)
results = []
results_unsorted = []
results_exact = []
results_start = []
results_other = []
x = 0
while url and x < API_MAX_REQUESTS:
log.debug(url)
r = requests.get(url)
if not r.status_code == 200:
return []
data = json.loads(r.text.replace('api.discogs.com', DISCOGS_HOST))
url = reduce(dict.get, ['pagination', 'urls', 'next'], data)
for r in data['results']:
if 'title' in r:
title = r['title']
formatted_title = re.sub(name_pattern, '', title)
r['index'] = get_index(title)
r['formatted_title'] = formatted_title
r['uri'] = 'https://www.discogs.com%s' % r['uri']
r['dist'] = distance(formatted_title.lower(), q_stripped.lower())
r['dist1'] = jaro(formatted_title.lower(), q_stripped.lower())
r['dist2'] = jaro_winkler(formatted_title.lower(), q_stripped.lower())
r['dist3'] = ratio(formatted_title.lower(), q_stripped.lower())
# print r['dist'],
# print r['dist1'],
# print r['dist2'],
# print r['dist3'],
# print formatted_title.lower(),
# print '::: {0} <> {1}'.format(formatted_title.lower(), q_stripped.lower())
results_unsorted.append(r)
if formatted_title.lower() == q_stripped.lower():
#print 'exact', formatted_title.lower()
results_exact.append(r)
elif formatted_title.lower().startswith(q_stripped.lower()[0:10]):
#print 'start', formatted_title.lower()
results_start.append(r)
else:
#print 'other', formatted_title.lower()
results_other.append(r)
x += 1
#results = sort_results(results_exact) + sort_results(results_start)+ sort_results(results_other)
results = sort_results_by_distance(results_unsorted)
if item_type == 'artist':
results = populate_results(results)
return results[0:limit]
