def name_signature(first, last):
ident = iunaccent(last.strip())
ident = nn_escaping_chars_re.sub('',ident)
ident = nn_final_nontext_re.sub('',ident)
ident = nn_nontext_re.sub('-',ident)
if len(first):
ident = iunaccent(first[0])+'-'+ident
return ident
def name_signature(first, last):
ident = iunaccent(last.strip())
ident = nn_escaping_chars_re.sub('', ident)
ident = nn_final_nontext_re.sub('', ident)
ident = nn_nontext_re.sub('-', ident)
if len(first):
ident = iunaccent(first[0]) + '-' + ident
return ident
def name_similarity(a, b):
"""
Returns a float: how similar are these two names?
"""
if not a or not b or len(a) != 2 or len(b) != 2:
return False
firstA, lastA = a
firstB, lastB = b
firstA = iunaccent(firstA)
firstB = iunaccent(firstB)
lastA = iunaccent(lastA)
lastB = iunaccent(lastB)
if lastA != lastB:
return 0.
partsA, sepsA = split_name_words(firstA)
partsB, sepsB = split_name_words(firstB)
parts = list(zip(partsA, partsB))
if not all(map(match_first_names, parts)):
# Try to match in reverse
partsA.reverse()
partsB.reverse()
parts = list(zip(partsA, partsB))
if not all(map(match_first_names, parts)):
return 0.
maxlen = max(len(partsA), len(partsB))
sumscores = 0
expanded = []
for i in range(maxlen):
if i < len(parts):
sumscores += weight_first_names(parts[i])
expanded.append((len(partsA[i]) > 1, len(partsB[i]) > 1))
elif i < len(partsA):
sumscores -= 0.25*weight_first_name(partsA[i])
expanded.append((len(partsA[i]) > 1, False))
else:
sumscores -= 0.25*weight_first_name(partsB[i])
expanded.append((False, len(partsB[i]) > 1))
# Make sure expanded first names of A are included in that of B
# or that of B and included in that of A
# This prevents ('Amanda P.','Brown') and ('A. Patrick','Brown')
# frow matching
if not (all([wa or not wb for wa, wb in expanded]) or
all([wb or not wa for wa, wb in expanded])):
return 0.
sumscores = max(min(sumscores, 1), 0)
return sumscores
def name_similarity(a, b):
"""
Returns a float: how similar are these two names?
Examples:
name_similarity(('Robin', 'Ryder'),('Robin', 'Ryder')) == 0.8
name_similarity(('Robin', 'Ryder'),('R.', 'Ryder')) == 0.4
name_similarity(('R.', 'Ryder'),('R.', 'Ryder')) == 0.4
name_similarity(('Robin J.', 'Ryder'),('R.', 'Ryder')) ==0.3
name_similarity(('Robin J.', 'Ryder'),('R. J.', 'Ryder')) == 0.8
name_similarity(('R. J.', 'Ryder'),('J.', 'Ryder')) == 0.3
name_similarity(('Robin', 'Ryder'),('Robin J.', 'Ryder')) == 0.7
name_similarity(('W. Timothy','Gowers'), ('Timothy','Gowers')) == 0.7
name_similarity(('Robin K.','Ryder'), ('Robin J.', 'Ryder')) == 0
name_similarity(('Claire', 'Mathieu'),('Claire', 'Kenyon-Mathieu') == 0
"""
if not a or not b:
return False
(firstA, lastA) = a
(firstB, lastB) = b
firstA = iunaccent(firstA)
firstB = iunaccent(firstB)
lastA = iunaccent(lastA)
lastB = iunaccent(lastB)
if lastA != lastB:
return 0.
#if firstA == firstB:
# return 1.
partsA, sepsA = split_name_words(firstA)
partsB, sepsB = split_name_words(firstB)
parts = zip(partsA, partsB)
if not all(map(match_first_names, parts)):
# Try to match in reverse
partsA.reverse()
partsB.reverse()
parts = zip(partsA, partsB)
if not all(map(match_first_names, parts)):
return 0.
maxlen = max(len(partsA), len(partsB))
sumscores = 0
for i in range(maxlen):
if i < len(parts):
sumscores += weight_first_names(parts[i])
elif i < len(partsA):
sumscores -= 0.25 * weight_first_name(partsA[i])
else:
sumscores -= 0.25 * weight_first_name(partsB[i])
sumscores = max(min(sumscores, 1), 0)
return sumscores
def name_similarity(a,b):
"""
Returns a float: how similar are these two names?
Examples:
name_similarity(('Robin', 'Ryder'),('Robin', 'Ryder')) == 0.8
name_similarity(('Robin', 'Ryder'),('R.', 'Ryder')) == 0.4
name_similarity(('R.', 'Ryder'),('R.', 'Ryder')) == 0.4
name_similarity(('Robin J.', 'Ryder'),('R.', 'Ryder')) ==0.3
name_similarity(('Robin J.', 'Ryder'),('R. J.', 'Ryder')) == 0.8
name_similarity(('R. J.', 'Ryder'),('J.', 'Ryder')) == 0.3
name_similarity(('Robin', 'Ryder'),('Robin J.', 'Ryder')) == 0.7
name_similarity(('W. Timothy','Gowers'), ('Timothy','Gowers')) == 0.7
name_similarity(('Robin K.','Ryder'), ('Robin J.', 'Ryder')) == 0
name_similarity(('Claire', 'Mathieu'),('Claire', 'Kenyon-Mathieu') == 0
"""
if not a or not b:
return False
(firstA,lastA) = a
(firstB,lastB) = b
firstA = iunaccent(firstA)
firstB = iunaccent(firstB)
lastA = iunaccent(lastA)
lastB = iunaccent(lastB)
if lastA != lastB:
return 0.
#if firstA == firstB:
# return 1.
partsA, sepsA = split_name_words(firstA)
partsB, sepsB = split_name_words(firstB)
parts = zip(partsA, partsB)
if not all(map(match_first_names, parts)):
# Try to match in reverse
partsA.reverse()
partsB.reverse()
parts = zip(partsA, partsB)
if not all(map(match_first_names, parts)):
return 0.
maxlen = max(len(partsA), len(partsB))
sumscores = 0
for i in range(maxlen):
if i < len(parts):
sumscores += weight_first_names(parts[i])
elif i < len(partsA):
sumscores -= 0.25*weight_first_name(partsA[i])
else:
sumscores -= 0.25*weight_first_name(partsB[i])
sumscores = max(min(sumscores, 1), 0)
return sumscores
def populate_identifiers(apps, se):
Institution = apps.get_model('papers', 'Institution')
for i in Institution.objects.all():
if i.country and i.name:
i.identifiers = [i.identifier, i.country + ':' + iunaccent(i.name)]
else:
i.identifiers = [i.identifier]
i.save(update_fields=['identifiers'])
def populate_identifiers(apps, se):
Institution = apps.get_model('papers', 'Institution')
for i in Institution.objects.all():
if i.country and i.name:
i.identifiers = [i.identifier, i.country+':'+iunaccent(i.name)]
else:
i.identifiers = [i.identifier]
i.save(update_fields=['identifiers'])
def create(cls, first, last):
"""
Creates an instance of the Name object without saving it.
Useful for name lookups where we are not sure we want to
keep the name in the model.
"""
instance = cls()
instance.first = sanitize_html(first[:MAX_NAME_LENGTH].strip())
instance.last = sanitize_html(last[:MAX_NAME_LENGTH].strip())
instance.full = iunaccent(instance.first+' '+instance.last)
return instance
def create(cls, first, last):
"""
Creates an instance of the Name object without saving it.
Useful for name lookups where we are not sure we want to
keep the name in the model.
"""
instance = cls()
instance.first = sanitize_html(first[:MAX_NAME_LENGTH].strip())
instance.last = sanitize_html(last[:MAX_NAME_LENGTH].strip())
instance.full = iunaccent(instance.first+' '+instance.last)
return instance
def shallower_name_similarity(a, b):
"""
Same as name_similarity, but accepts differences in the last names.
This heuristics is more costly but is only used to attribute an ORCID
affiliation to the right author in papers fetched from ORCID.
(in the next function)
"""
if not a or not b or len(a) != 2 or len(b) != 2:
return False
firstA, lastA = a
firstB, lastB = b
# Matching last names
lastA = iunaccent(lastA)
lastB = iunaccent(lastB)
wordsA, _ = split_name_words(lastA)
wordsB, _ = split_name_words(lastB)
wordsA = set(wordsA)
wordsB = set(wordsB)
if not wordsA or not wordsB:
return False
ratio = float(len(wordsA & wordsB)) / len(wordsA | wordsB)
partsA, sepsA = split_name_words(firstA)
partsB, sepsB = split_name_words(firstB)
partsA = [p[0] for p in partsA]
partsB = [p[0] for p in partsB]
parts = zip(partsA, partsB)
if not all(map(match_first_names, parts)):
# Try to match in reverse
partsA.reverse()
partsB.reverse()
parts = zip(partsA, partsB)
if not all(map(match_first_names, parts)):
return 0.
maxlen = max(len(partsA), len(partsB))
return ratio * (len(parts) + 1) / (maxlen + 1)
def shallower_name_similarity(a, b):
"""
Same as name_similarity, but accepts differences in the last names.
This heuristics is more costly but is only used to attribute an ORCID
affiliation to the right author in papers fetched from ORCID.
(in the next function)
"""
if not a or not b or len(a) != 2 or len(b) != 2:
return False
firstA, lastA = a
firstB, lastB = b
# Matching last names
lastA = iunaccent(lastA)
lastB = iunaccent(lastB)
wordsA, _ = split_name_words(lastA)
wordsB, _ = split_name_words(lastB)
wordsA = set(wordsA)
wordsB = set(wordsB)
if not wordsA or not wordsB:
return False
ratio = float(len(wordsA & wordsB)) / len(wordsA | wordsB)
partsA, sepsA = split_name_words(firstA)
partsB, sepsB = split_name_words(firstB)
partsA = [ p[0] for p in partsA ]
partsB = [ p[0] for p in partsB ]
parts = list(zip(partsA, partsB))
if not all(map(match_first_names, parts)):
# Try to match in reverse
partsA.reverse()
partsB.reverse()
parts = list(zip(partsA, partsB))
if not all(map(match_first_names, parts)):
return 0.
maxlen = max(len(partsA), len(partsB))
return ratio*(len(parts)+1)/(maxlen+1)
def shallower_name_similarity(a, b):
"""
Same as name_similarity, but accepts differences in the last names.
This heuristics is more costly but is only used to attribute an ORCID
affiliation to the right author in papers fetched from ORCID.
"""
if not a or not b:
return False
firstA, lastA = a
firstB, lastB = b
# Matching last names
lastA = iunaccent(lastA)
lastB = iunaccent(lastB)
wordsA, sepA = split_name_words(lastA)
wordsB, sepB = split_name_words(lastB)
wordsA = set(wordsA)
wordsB = set(wordsB)
ratio = float(len(wordsA & wordsB)) / len(wordsA | wordsB)
partsA, sepsA = split_name_words(firstA)
partsB, sepsB = split_name_words(firstB)
partsA = map(lambda x: x[0], partsA)
partsB = map(lambda x: x[0], partsB)
parts = zip(partsA, partsB)
if not all(map(match_first_names, parts)):
# Try to match in reverse
partsA.reverse()
partsB.reverse()
parts = zip(partsA, partsB)
if not all(map(match_first_names, parts)):
return 0.
maxlen = max(len(partsA), len(partsB))
if maxlen > 0:
return ratio * len(parts) / maxlen
return 0.
def shallower_name_similarity(a, b):
"""
Same as name_similarity, but accepts differences in the last names.
This heuristics is more costly but is only used to attribute an ORCID
affiliation to the right author in papers fetched from ORCID.
"""
if not a or not b:
return False
firstA, lastA = a
firstB, lastB = b
# Matching last names
lastA = iunaccent(lastA)
lastB = iunaccent(lastB)
wordsA, sepA = split_name_words(lastA)
wordsB, sepB = split_name_words(lastB)
wordsA = set(wordsA)
wordsB = set(wordsB)
ratio = float(len(wordsA & wordsB)) / len(wordsA | wordsB)
partsA, sepsA = split_name_words(firstA)
partsB, sepsB = split_name_words(firstB)
partsA = map(lambda x: x[0], partsA)
partsB = map(lambda x: x[0], partsB)
parts = zip(partsA, partsB)
if not all(map(match_first_names, parts)):
# Try to match in reverse
partsA.reverse()
partsB.reverse()
parts = zip(partsA, partsB)
if not all(map(match_first_names, parts)):
return 0.
maxlen = max(len(partsA), len(partsB))
if maxlen > 0:
return ratio*len(parts)/maxlen
return 0.
def test_iunaccent(self):
self.assertEqual(iunaccent('BÉPO forever'), 'bepo forever')
def name_similarity(a, b):
"""
Returns a float: how similar are these two names?
Examples:
>>> int(10*name_similarity(('Robin', 'Ryder'),('Robin', 'Ryder')))
8
>>> int(10*name_similarity(('Robin', 'Ryder'),('R.', 'Ryder')))
4
>>> int(10*name_similarity(('R.', 'Ryder'),('R.', 'Ryder')))
4
>>> int(10*name_similarity(('Robin J.', 'Ryder'),('R.', 'Ryder')))
3
>>> int(10*name_similarity(('Robin J.', 'Ryder'),('R. J.', 'Ryder')))
8
>>> int(10*name_similarity(('R. J.', 'Ryder'),('J.', 'Ryder')))
3
>>> int(10*name_similarity(('Robin', 'Ryder'),('Robin J.', 'Ryder')))
7
>>> int(10*name_similarity(('W. Timothy','Gowers'), ('Timothy','Gowers')))
7
>>> int(10*name_similarity(('Robin K.','Ryder'), ('Robin J.', 'Ryder')))
0
>>> int(10*name_similarity(('Claire', 'Mathieu'),('Claire', 'Kenyon-Mathieu')))
0
>>> int(10*name_similarity(('Amanda P.','Brown'),('Patrick','Brown')))
0
"""
if not a or not b:
return False
(firstA, lastA) = a
(firstB, lastB) = b
firstA = iunaccent(firstA)
firstB = iunaccent(firstB)
lastA = iunaccent(lastA)
lastB = iunaccent(lastB)
if lastA != lastB:
return 0.
partsA, sepsA = split_name_words(firstA)
partsB, sepsB = split_name_words(firstB)
parts = zip(partsA, partsB)
if not all(map(match_first_names, parts)):
# Try to match in reverse
partsA.reverse()
partsB.reverse()
parts = zip(partsA, partsB)
if not all(map(match_first_names, parts)):
return 0.
maxlen = max(len(partsA), len(partsB))
sumscores = 0
expanded = []
for i in range(maxlen):
if i < len(parts):
sumscores += weight_first_names(parts[i])
expanded.append((len(partsA[i]) > 1, len(partsB[i]) > 1))
elif i < len(partsA):
sumscores -= 0.25 * weight_first_name(partsA[i])
expanded.append((len(partsA[i]) > 1, False))
else:
sumscores -= 0.25 * weight_first_name(partsB[i])
expanded.append((False, len(partsB[i]) > 1))
# Make sure expanded first names of A are included in that of B
# or that of B and included in that of A
# This prevents ('Amanda P.','Brown') and ('A. Patrick','Brown')
# frow matching
if not (all([a or not b for a, b in expanded])
or all([b or not a for a, b in expanded])):
return 0.
sumscores = max(min(sumscores, 1), 0)
return sumscores
def name_similarity(a,b):
"""
Returns a float: how similar are these two names?
Examples:
>>> int(10*name_similarity(('Robin', 'Ryder'),('Robin', 'Ryder')))
8
>>> int(10*name_similarity(('Robin', 'Ryder'),('R.', 'Ryder')))
4
>>> int(10*name_similarity(('R.', 'Ryder'),('R.', 'Ryder')))
4
>>> int(10*name_similarity(('Robin J.', 'Ryder'),('R.', 'Ryder')))
3
>>> int(10*name_similarity(('Robin J.', 'Ryder'),('R. J.', 'Ryder')))
8
>>> int(10*name_similarity(('R. J.', 'Ryder'),('J.', 'Ryder')))
3
>>> int(10*name_similarity(('Robin', 'Ryder'),('Robin J.', 'Ryder')))
7
>>> int(10*name_similarity(('W. Timothy','Gowers'), ('Timothy','Gowers')))
7
>>> int(10*name_similarity(('Robin K.','Ryder'), ('Robin J.', 'Ryder')))
0
>>> int(10*name_similarity(('Claire', 'Mathieu'),('Claire', 'Kenyon-Mathieu')))
0
>>> int(10*name_similarity(('Amanda P.','Brown'),('Patrick','Brown')))
0
"""
if not a or not b:
return False
(firstA,lastA) = a
(firstB,lastB) = b
firstA = iunaccent(firstA)
firstB = iunaccent(firstB)
lastA = iunaccent(lastA)
lastB = iunaccent(lastB)
if lastA != lastB:
return 0.
partsA, sepsA = split_name_words(firstA)
partsB, sepsB = split_name_words(firstB)
parts = zip(partsA, partsB)
if not all(map(match_first_names, parts)):
# Try to match in reverse
partsA.reverse()
partsB.reverse()
parts = zip(partsA, partsB)
if not all(map(match_first_names, parts)):
return 0.
maxlen = max(len(partsA), len(partsB))
sumscores = 0
expanded = []
for i in range(maxlen):
if i < len(parts):
sumscores += weight_first_names(parts[i])
expanded.append((len(partsA[i])>1, len(partsB[i])>1))
elif i < len(partsA):
sumscores -= 0.25*weight_first_name(partsA[i])
expanded.append((len(partsA[i])>1, False))
else:
sumscores -= 0.25*weight_first_name(partsB[i])
expanded.append((False, len(partsB[i])>1))
# Make sure expanded first names of A are included in that of B
# or that of B and included in that of A
# This prevents ('Amanda P.','Brown') and ('A. Patrick','Brown')
# frow matching
if not (all([a or not b for a,b in expanded]) or
all([b or not a for a,b in expanded])):
return 0.
sumscores = max(min(sumscores, 1), 0)
return sumscores
