def __init__( self ):
self.ling = LinguisticUtility()
class EntityExtract():
def __init__( self ):
self.ling = LinguisticUtility()
def extract_name( self, text ):
text = self.ling.remove_diacritics( text )
text = text.replace( '\n', ' newline ' )
tokenized = self.ling.tokenizer( text )
tagged = nltk.pos_tag(tokenized)
entities = nltk.ne_chunk(tagged, binary=True )
nodes = self.getNodes( entities, 'NE' )
freq = dict();
for n in nodes :
if n in freq :
freq[n] = freq[n] + 1
else:
freq[n] = 1
freq = self.merge_name( freq )
return freq
def getNodes( self, node, node_type ):
res = []
if( hasattr( node, 'label' ) and node.label ):
if node.label() == node_type:
res.append(' '.join([child[0] for child in node]))
else:
for child in node:
res.extend( self.getNodes( child, node_type ) )
return res
def merge_name( self, names_dic ):
res = dict()
uniq_name = dict()
for k in names_dic.keys():
n_k = self.ling.remove_diacritics(k.lower())
if( n_k in uniq_name.keys() ):
old_key = uniq_name[n_k]
new_key = self.ling.choose_furthest_term( n_k, old_key , k )
if( new_key == k ):
old_freq = res.pop( uniq_name[n_k] )
res[new_key] = names_dic[k] + old_freq
uniq_name[n_k] = new_key
else:
res[old_key] = res[old_key] + names_dic[k]
else:
res[k] = names_dic[k]
uniq_name[n_k] = k
return res
