def test_singleton(self):
excpt_thrown = False
try:
Cogroo()
except TypeError, e:
excpt_thrown = True
self.assertEquals(Cogroo.MSG_CALL_INSTANCE, '%s' % e)
def test_lemmatize():
cogroo = Cogroo.Instance()
phrase_to_lemmatize = 'o entendimento das metas propostas oferece uma interessante oportunidade para verificação ' \
'do impacto na agilidade decisória '
expected_result = 'o entender de o meta propor oferecer um interessante oportunidade para verificação de o ' \
'impacto em o agilidade decisório'
assert expected_result == cogroo.lemmatize(phrase_to_lemmatize)
class CogrooSemanticizer:
cogroo = Cogroo.Instance()
entities_list = []
def __init__(self, text):
self.input_text = text
self.pos_tagged_text = self.cogroo.analyze(text).sentences[0]
def get_entities(self):
self.entities_list = self.search_chunks()
agglutinator = Agglutinator.Agglutinator(self.input_text,
self.entities_list)
self.entities_list = agglutinator.agglutinate()
self.clean_entities()
print("\n", "-" * 20, "> CogrooSemanticizer")
print(self.pos_tagged_text.chunks)
for entity in self.entities_list:
print(entity)
return self.entities_list
def search_chunks(self):
for chunk in self.pos_tagged_text.chunks:
if chunk.tag in n_tags or chunk.tag in v_tags:
self.entities_list = self.filter_chunk(chunk)
return self.entities_list
def filter_chunk(self, chunk):
'''
Retira expressões das frases
:param chunk:
:param self.entities_list:
:return: entities_list:
'''
for token in chunk.tokens:
if token.pos in n_tags:
entity = ec.Entity(text=token.lexeme,
start=token.start,
end=token.end,
tag='NP',
pos=token.pos)
self.entities_list.append(entity)
if token.pos in v_tags and token.lexeme == 'vem':
entity = ec.Entity(text=token.lexeme,
start=token.start,
end=token.end,
tag='VP',
pos=token.pos)
self.entities_list.append(entity)
return self.entities_list
def clean_entities(self):
for entity in self.entities_list:
text = entity.text.replace('_', ' ')
entity.text = text
def main(name):
file = open(name, mode='r')
cogroo = Cogroo.Instance()
sentences = []
for idx, line in enumerate(file):
aux = line.strip().split("\t")
frase = aux[0]
sentence = aux[0].replace("'", '"').replace(":", "#").replace(
";", "$").replace(".", "%")
en1 = aux[2].replace("=", "_").replace(".", "%")
en2 = aux[5].replace("=", "_").replace(".", "%")
if int(aux[1].split(",")[0]) > int(aux[4].split(",")[0]):
small = en2
big = en1
else:
small = en1
big = en2
sentence = sentence.replace(en1, "en1").replace(en2, "en2")
aux_features = []
pos = ""
check_en = 0
aux_lexeme = []
analyzer = cogroo.analyze(sentence).sentences[0].tokens
words = []
for idx, t in enumerate(analyzer):
if check_en == 0 and (t.lexeme == "en1" or t.lexeme == "en2"):
check_en = 1
aux_lexeme.append(t.lexeme)
t.lemma = small.replace("%", ".")
t.lexeme = small.replace("%", ".")
continue
elif check_en == 1 and (t.lexeme == "en1" or t.lexeme == "en2"):
if not t.lexeme == aux_lexeme[0]:
t.lemma = big.replace("%", ".")
t.lexeme = big.replace("%", ".")
break
elif check_en == 0:
continue
aux = t.synchunk if len(t.synchunk) <= 2 else t.synchunk[2:]
pos += " " + t.pos
vector = [t.lemma, t.pos, t.chunk[2:], aux]
vector.append("1") if t.pos[0] == "v" else vector.append("0")
vector.append("1") if t.pos == "adv" else vector.append("0")
# POS, Lemma and Syntactic Tags (-2, -1, 1, 2)
gap(analyzer, vector, idx, -2)
gap(analyzer, vector, idx, -1)
gap(analyzer, vector, idx, 1)
gap(analyzer, vector, idx, 2)
consecutive(analyzer, vector, idx, -2, -1)
consecutive(analyzer, vector, idx, -1, 0)
consecutive(analyzer, vector, idx, 0, 1)
consecutive(analyzer, vector, idx, 1, 2)
# Pattern based features
try:
vector.append("1") if analyzer[
idx - 2].pos[0] == "v" else vector.append("0")
except:
vector.append("0")
try:
vector.append("1") if analyzer[
idx - 1].pos[0] == "v" else vector.append("0")
except:
vector.append("0")
try:
vector.append("1") if analyzer[
idx + 1].pos[0] == "v" else vector.append("0")
except:
vector.append("0")
try:
vector.append("1") if analyzer[
idx + 2].pos[0] == "v" else vector.append("0")
except:
vector.append("0")
try:
vector.append("1") if analyzer[
idx +
1].pos == "prp" and t.pos[0] == "v" else vector.append("0")
except:
vector.append("0")
try:
vector.append("1") if analyzer[
idx +
1].pos == "art" and t.pos[0] == "v" else vector.append("0")
except:
vector.append("0")
try:
vector.append(
"1") if analyzer[idx + 2].pos == "art" and analyzer[
idx +
1].pos == "prp" and t.pos[0] == "v" else vector.append(
"0")
except:
vector.append("0")
try:
vector.append("1") if analyzer[
idx +
1].pos == "prp" and t.pos == "n" else vector.append("0")
except:
vector.append("0")
try:
vector.append("1") if analyzer[
idx +
1].pos == "prp" and t.pos == "adv" else vector.append("0")
except:
vector.append("0")
try:
vector.append(
"1") if analyzer[idx + 2].pos == "art" and analyzer[
idx +
1].pos == "prp" and t.pos == "adv" else vector.append(
"0")
except:
vector.append("0")
# Syntactic features
# Núcleo do sintagma
if not t.lexeme in ("o", "a", "os", "as") and t.pos in (
"n", "prop", "pron-det", "pron-pers",
"pron-indp") and "NP" in t.chunk:
vector.append("1")
else:
vector.append("0")
# Objeto Direto
if aux == "ACC": vector.append("1")
else: vector.append("0")
vector.insert(0, t.lexeme)
aux = line.strip().split("\t")
if aux[3] == "PER" or aux[6] == "PER": vector.append("O-PER")
elif aux[3] == "PLC" or aux[6] == "PLC": vector.append("O-PLC")
else: vector.append("O-O")
aux_features.append(vector)
for x in aux_features:
x.append(pos[1:])
x.append(str(len(pos[1:].split(" "))))
words.append(x)
en1 = en1.replace("%", ".")
en2 = en2.replace("%", ".")
sentences.append([words, frase, en1, en2])
# SALVA OS VETORES DE FEATURES
print(sentences)
with gzip.open('4_RelP/features_teste.txt.gz', 'wb') as f:
pickle.dump(sentences, f)
import csv
from cogroo_interface import Cogroo
cogroo = Cogroo.Instance()
classe = []
resposta = []
lema_list = []
analise = []
classes_pergs = []
i=0
with open('corpus.csv') as csvfile:
readCSV = csv.reader(csvfile, delimiter=',')
next(readCSV)
for row in readCSV:
classe.append(row[3])
resposta.append(row[2])
pergunta_lema = str(cogroo.lemmatize(row[1]))
pergunta_lema = pergunta_lema.lower() #Passa todas as palavras para lowercase
lema_list.append( str(pergunta_lema).split(' ') ) #adiciona as perguntas lematizadas em uma lista
#analise_perg = cogroo.analyze(pergunta_lema)
#analise.append(analise_perg.sentences[0].tokens) #guarda a analise morfologica das perguntas em uma lista
y=0
while(y<len(lema_list[i])):
perg = str(lema_list[i][y])
if(perg=='o' or perg=='de' or perg==',' or perg=='que' or perg=='qual' or perg=='a' or perg=='um' or perg=='.' or perg=='o('
or perg=='?' or perg=='' or perg=='ser' or perg=='quem' or perg=='em' or perg=='por' or perg=='algoritmo'
or perg=='ir' or perg=='se' or perg=='random_forest' or perg=='random' or perg=='forest') or perg=='para':
lema_list[i].pop(y)
y=y-1
def setUp(self):
self.cogroo = Cogroo.instance()
def __init__(self, plan):
self.cogroo = Cogroo.Instance()
self.plan = plan
def main(name):
file = open(name, "r")
cogroo = Cogroo.Instance()
sentences = []
for idx, line in enumerate(file):
aux = line.strip().split("\t")
'''print("Frase original -",aux[0])
print("EN1 -", aux[2])
print("EN2 -",aux[7])'''
# PRÉ-PROCESSAMENTO
frase = aux[0]
sentence = aux[0].replace("'", '"').replace(":", "#").replace(
";", "$").replace(".", "%")
en1 = aux[2].replace("=", "_").replace(".", "%")
rel_num = aux[4].split(",")
rel = aux[5].split(" ")
en2 = aux[7].replace("=", "_").replace(".", "%")
if rel_num[0] == 'None': rel_start = 0
else: rel_start = int(rel_num[0])
if int(aux[1].split(",")[0]) > int(aux[6].split(",")[0]):
small = en2
big = en1
else:
small = en1
big = en2
sentence = sentence.replace(en1, "en1").replace(en2, "en2")
'''print("Frase modificada -",sentence, "\n")
ajuda = "Sem relação" if rel == [''] else " ".join(rel)
print(ajuda, "\n")'''
aux_rel = []
aux_features = []
aux_lexeme = []
pos = ""
check_en = 0
count_rel = 0
analyzer = cogroo.analyze(sentence).sentences[0].tokens
words = []
for idx, t in enumerate(analyzer):
if check_en == 0 and (t.lexeme == "en1" or t.lexeme == "en2"):
check_en = 1
aux_lexeme.append(t.lexeme)
t.lemma = small.replace("%", ".")
t.lexeme = small.replace("%", ".")
continue
elif check_en == 1 and (t.lexeme == "en1" or t.lexeme == "en2"):
if not t.lexeme == aux_lexeme[0]:
t.lemma = big.replace("%", ".")
t.lexeme = big.replace("%", ".")
break
elif check_en == 0:
continue
# CLASSE
if count_rel == len(rel): classification = 0
elif t.lexeme == rel[count_rel]:
classification = 1
count_rel += 1
else:
classification = 0
# FEATURES
aux = t.synchunk if len(t.synchunk) <= 2 else t.synchunk[2:]
pos += " " + t.pos
vector = [t.lemma, t.pos, t.chunk[2:], aux]
vector.append("1") if t.pos[0] == "v" else vector.append("0")
vector.append("1") if t.pos == "adv" else vector.append("0")
# POS, Lemma and Syntactic Tags (-2, -1, 1, 2)
gap(analyzer, vector, idx, -2)
gap(analyzer, vector, idx, -1)
gap(analyzer, vector, idx, 1)
gap(analyzer, vector, idx, 2)
consecutive(analyzer, vector, idx, -2, -1)
consecutive(analyzer, vector, idx, -1, 0)
consecutive(analyzer, vector, idx, 0, 1)
consecutive(analyzer, vector, idx, 1, 2)
# Pattern based features
try:
vector.append("1") if analyzer[
idx - 2].pos[0] == "v" else vector.append("0")
except:
vector.append("0")
try:
vector.append("1") if analyzer[
idx - 1].pos[0] == "v" else vector.append("0")
except:
vector.append("0")
try:
vector.append("1") if analyzer[
idx + 1].pos[0] == "v" else vector.append("0")
except:
vector.append("0")
try:
vector.append("1") if analyzer[
idx + 2].pos[0] == "v" else vector.append("0")
except:
vector.append("0")
try:
vector.append("1") if analyzer[
idx +
1].pos == "prp" and t.pos[0] == "v" else vector.append("0")
except:
vector.append("0")
try:
vector.append("1") if analyzer[
idx +
1].pos == "art" and t.pos[0] == "v" else vector.append("0")
except:
vector.append("0")
try:
vector.append(
"1") if analyzer[idx + 2].pos == "art" and analyzer[
idx +
1].pos == "prp" and t.pos[0] == "v" else vector.append(
"0")
except:
vector.append("0")
try:
vector.append("1") if analyzer[
idx +
1].pos == "prp" and t.pos == "n" else vector.append("0")
except:
vector.append("0")
try:
vector.append("1") if analyzer[
idx +
1].pos == "prp" and t.pos == "adv" else vector.append("0")
except:
vector.append("0")
try:
vector.append(
"1") if analyzer[idx + 2].pos == "art" and analyzer[
idx +
1].pos == "prp" and t.pos == "adv" else vector.append(
"0")
except:
vector.append("0")
# Syntactic features
# Núcleo do sintagma
if not t.lexeme in ("o", "a", "os", "as") and t.pos in (
"n", "prop", "pron-det", "pron-pers",
"pron-indp") and "NP" in t.chunk:
vector.append("1")
else:
vector.append("0")
# Objeto Direto
if aux == "ACC": vector.append("1")
else: vector.append("0")
vector.insert(0, t.lexeme)
vector.insert(0, str(classification))
aux = line.strip().split("\t")
if aux[3] == "PER" or aux[8] == "PER": vector.append("O-PER")
elif aux[3] == "PLC" or aux[8] == "PLC": vector.append("O-PLC")
else: vector.append("O-O")
aux_features.append(vector)
'''print(classification, t.lexeme)'''
for x in aux_features:
x.append(pos[1:])
x.append(str(len(pos[1:].split(" "))))
words.append(x)
'''print("\n")'''
en1 = en1.replace("%", ".")
en2 = en2.replace("%", ".")
sentences.append([words, frase, en1, en2])
# SALVA OS VETORES DE FEATURES
with gzip.open('features_treino.txt.gz', 'wb') as f:
pickle.dump(sentences, f)
def __init__(self):
self.cogroo = Cogroo.Instance()
