def __init__(self, lang):
if lang == LANG.DE:
self.spre = ShortestPathRE().de_lang()
self.pbre = PatternBasedRE().de_lang()
#self.entity_extractor = FlairEntityExtractor().de_lang()
self.entity_extractor = SpacyEntityExtractor().de_lang()
else:
self.spre = ShortestPathRE().en_lang()
self.pbre = PatternBasedRE().en_lang()
#self.entity_extractor = FlairEntityExtractor().en_lang()
self.entity_extractor = SpacyEntityExtractor().en_lang()
class RelationExtractor:
def __init__(self, lang):
if lang == LANG.DE:
self.spre = ShortestPathRE().de_lang()
self.pbre = PatternBasedRE().de_lang()
#self.entity_extractor = FlairEntityExtractor().de_lang()
self.entity_extractor = SpacyEntityExtractor().de_lang()
else:
self.spre = ShortestPathRE().en_lang()
self.pbre = PatternBasedRE().en_lang()
#self.entity_extractor = FlairEntityExtractor().en_lang()
self.entity_extractor = SpacyEntityExtractor().en_lang()
def extract_relations(self, text, plot_graph=False, validate=False, out_val_file=None):
extracted_relations = []
for sentence in sent_tokenize(text):
entities, per_entities = self.entity_extractor.extract_entities(sentence)
logger.debug(f'Extracted entities: {entities}')
# Shortest path relation extraction
if len(per_entities) > 0: # PER-PER or USR-PER
extracted_relations = self.spre.extract_sp_relation(entities, per_entities, sentence, plot_graph)
# Pattern based relation extraction
else: # USR-REL
extracted_relations = self.pbre.extract_rel(sentence)
if validate:
with open(out_val_file,
'a', encoding='utf-8') as f:
validated = f'{extracted_relations}; {sentence}\n'
f.write(validated)
return extracted_relations
def test_extract_rel_8():
utterance = u'''sadly i am on call with my uncle in the hospital so i never let mine die'''
pbre = PatternBasedRE.en_lang()
result = pbre.extract_rel(utterance, plot_tree=False)
assert result == [('uncle-of', 'USER')]
def test_extract_rel_6():
utterance = 'i miss my wife and kids so much'
pbre = PatternBasedRE().en_lang()
result = pbre.extract_rel(utterance, plot_tree=False)
assert result == [('wife-of', 'USER')]
def test_extract_rel_7():
utterance = 'no , my dad taught me good music and good work ethics.'
pbre = PatternBasedRE.en_lang()
result = pbre.extract_rel(utterance, plot_tree=False)
assert result == [('father-of', 'USER')]
def test_extract_rel_5():
utterance = 'Ich habe einen Bruder'
pbre = PatternBasedRE().de_lang()
result = pbre.extract_rel(utterance, plot_tree=False)
assert result == [('brother-of', 'USER')]
def test_extract_rel_1():
utterance = 'I have older brother who lives in Berlin'
pbre = PatternBasedRE().en_lang()
result = pbre.extract_rel(utterance, plot_tree=False)
assert result == [('brother-of', 'USER')]
def test_extract_rel_4():
utterance = 'My little sister Lisa is moving to London'
pbre = PatternBasedRE().en_lang()
result = pbre.extract_rel(utterance, plot_tree=True)
assert result == [('sister-of', 'USER')]
def test_extract_rel_3():
utterance = 'I have one brother'
pbre = PatternBasedRE().en_lang()
result = pbre.extract_rel(utterance, plot_tree=True)
assert result == [('brother-of', 'USER')]
def test_extract_rel_2():
utterance = 'I have two sisters'
pbre = PatternBasedRE().en_lang()
result = pbre.extract_rel(utterance, plot_tree=True)
assert result == [('sister-of', 'USER')]