def ident(self, name):
orig_name = name
best_matches = []
results = self.entity_type.query.get_results()
mapper = self.entity_type.result_mapper
name = mapper.stand(name)
for uri, given, stand in mapper.map(results):
dist = jaccard_ngram_dist(stand,name,3)
best_matches.append(((given,uri),dist))
if len(best_matches)>100:
best_matches.sort(key=lambda x:x[1])
best_matches = best_matches[:self.max_matches]
best_matches.sort(key=lambda x:x[1])
best_matches = best_matches[:self.max_matches]
best_dist = best_matches[0][1]
possibles = [best_matches[0][0]]
for match, dist in best_matches[1:]:
if (dist+best_dist)==0:
percent_diff = 0
else:
percent_diff = (dist-best_dist)*2/float(dist+best_dist)
if percent_diff < self.max_percent_diff:
possibles.append(match)
if len(possibles)>1:
identified = self.prompt_possibles(orig_name, possibles)
else:
identified = possibles[0]
return identified
'adam smith',
'bob smith',
'carl smith',
'dale jones',
'ernest kirstein'
]
import ngrams
to_compare = [
'tom smith',
'john smith',
'tom'
]
#to_compare = ['rj smith', 'rj', 'cs smith']
def run_comps(f, to_compare):
for i,s1 in enumerate(to_compare):
for s2 in to_compare[i+1:]:
print f(s1,s2), repr(s1), repr(s2)
print "Without Info:"
d = lambda a,b: ngrams.jaccard_ngram_dist(a,b,3)
run_comps(d, to_compare)
print "\nSample Set:"
print '\t'+'\n\t'.join(names)
model = ngram_model(names, 3)
print "\nWith Info:"
d = lambda a,b: info_ngram_dist(3,a,b,model)
run_comps(d, to_compare)