proNlp1/newsTrain.py

383 lines
9.5 KiB
Python
Raw Normal View History

2017-11-28 06:52:47 +00:00
from sklearn.feature_extraction.text import TfidfVectorizer
from stopWords import stopWrdList
from retEmoDict import emoDic
from clust import clustering
2017-11-28 01:14:38 +00:00
2017-11-28 06:52:47 +00:00
def trainPre(word_array, dict):
2017-11-28 01:14:38 +00:00
default = 'NA'
alegria = []
enojo = []
miedo = []
repulsion = []
sorpresa = []
tristeza = []
proper = []
part = []
for word in word_array:
if dict.get(str(word), default) == 'Alegría':
alegria.append(1)
proper.append(word)
if dict.get(str(word), default) == 'Enojo':
enojo.append(1)
proper.append(word)
if dict.get(str(word), default) == 'Miedo':
miedo.append(1)
proper.append(word)
if dict.get(str(word), default) == 'Repulsión':
repulsion.append(1)
proper.append(word)
if dict.get(str(word), default) == 'Sorpresa':
sorpresa.append(1)
proper.append(word)
if dict.get(str(word), default) == 'Tristeza':
tristeza.append(1)
proper.append(word)
if dict.get(str(word), default) == 'Positivo':
part.append('PRI')
proper.append(word)
if dict.get(str(word), default) == 'Negativo':
part.append('CONTRA')
proper.append(word)
if dict.get(str(word), default) == 'Neutro':
part.append('NEU')
proper.append(word)
if dict.get(str(word), default) == 'NA':
proper.append(word)
part = set(part)
flag = list(part)
vect = set(proper)
vect = list(vect)
tot = len(word_array)
alegria = sum(alegria)
enojo = sum(enojo)
miedo = sum(miedo)
repulsion = sum(repulsion)
sorpresa = sum(sorpresa)
tristeza = sum(tristeza)
pos = (alegria + sorpresa) / tot
neg = (enojo + miedo + repulsion + tristeza) / tot
if len(flag) == 0:
flag = ['NEU']
2017-11-28 06:52:47 +00:00
return [pos, neg, flag, vect]
def corporizer():
emoDict = emoDic()
clust = clustering()
temp = []
for i in clust:
temp.append(trainPre(i, emoDict))
tempy = []
for vect in temp:
tempy.append(' '.join(vect[3]))
return tempy
def flagger():
emoDict = emoDic()
clust = clustering()
temp = []
for i in clust:
temp.append(trainPre(i, emoDict))
flag = []
for j in temp:
#print(j[2])
if j[2] == (['CONTRA', 'NEU', 'PRI'] or ['NEU', 'CONTRA', 'PRI'] or ['NEU', 'PRI', 'CONTRA'] or
['PRI', 'NEU', 'CONTRA'] or ['CONTRA', 'PRI', 'NEU'] or ['PRI', 'CONTRA', 'NEU']):
flag.append(1)
#else:
# flag.append(0)
if j[2] == (['CONTRA', 'PRI'] or ['PRI', 'CONTRA']):
flag.append(1)
#else:
# flag.append(6)
if j[2] == ['NEU']:
flag.append(1)
#else:
# flag.append(7)
if j[2] == (['PRI'] or ['NEU', 'PRI'] or ['PRI', 'NEU']):
flag.append(2)
#else:
# flag.append(8)
if j[2] == (['CONTRA'] or ['NEU', 'CONTRA'] or ['CONTRA', 'NEU']):
flag.append(3)
#else:
# flag.append(9)
index = []
for i in temp:
if i[0] == i[1]:
index.append(1)
if i[0] > i[1]:
index.append(2)
if i[0] < i[1]:
index.append(3)
lenFlag = len(flag)
lenInde = len(index)
if lenFlag < lenInde:
for i in range(lenInde - lenFlag):
flag.append(1)
return (index, flag)
def operate_on_Narray(A, B, function):
try:
return [operate_on_Narray(a, b, function) for a, b in zip(A, B)]
except TypeError as e:
# Not iterable
return function(A, B)
def trainVect():
flag = flagger()
corpus = corporizer()
stop_words = stopWrdList()
2017-12-04 18:01:46 +00:00
vectorizer = TfidfVectorizer(strip_accents='ascii', analyzer='word', stop_words=stop_words, max_features=100)
2017-11-28 06:52:47 +00:00
X = vectorizer.fit_transform(corpus)
vector = X.toarray()
long = len(flag[0])
part_neu_ind = []
part_neg_ind = []
part_pos_ind = []
cont_neu_ind = []
cont_neg_ind = []
cont_pos_ind = []
neut_neu_ind = []
neut_neg_ind = []
neut_pos_ind = []
# flag 0 has emotion info, flag 1 has political party info
# 1 is neutral emo ; 2 is positive emo ; 3 is negative emo
# 1 is neutral ; 2 is pol; 3 is opposition
for s in range(long):
if flag[0][s] == 1 and flag[1][s] == 1:
neut_neu_ind.append(s)
if flag[0][s] == 1 and flag[1][s] == 2:
part_neu_ind.append(s)
if flag[0][s] == 1 and flag[1][s] == 3:
cont_neu_ind.append(s)
if flag[0][s] == 2 and flag[1][s] == 2:
part_pos_ind.append(s)
if flag[0][s] == 2 and flag[1][s] == 3:
cont_pos_ind.append(s)
if flag[0][s] == 2 and flag[1][s] == 1:
neut_pos_ind.append(s)
if flag[0][s] == 3 and flag[1][s] == 1:
neut_neg_ind.append(s)
if flag[0][s] == 3 and flag[1][s] == 2:
part_neg_ind.append(s)
if flag[0][s] == 3 and flag[1][s] == 3:
cont_neg_ind.append(s)
part_neu_vect = [vector[x] for x in part_neu_ind]
part_neg_vect = [vector[x] for x in part_neg_ind]
part_pos_vect = [vector[x] for x in part_pos_ind]
cont_neu_vect = [vector[x] for x in cont_neu_ind]
cont_neg_vect = [vector[x] for x in cont_neg_ind]
cont_pos_vect = [vector[x] for x in cont_pos_ind]
neut_neu_vect = [vector[x] for x in neut_neu_ind]
neut_neg_vect = [vector[x] for x in neut_neg_ind]
neut_pos_vect = [vector[x] for x in neut_pos_ind]
2017-12-04 18:01:46 +00:00
############################################ 1
2017-11-28 06:52:47 +00:00
len1 = len(part_neu_vect)
if len1 != 0:
for a in range(len1):
tmp = part_neu_vect[0]
tmp = operate_on_Narray(part_neu_vect[0], tmp[a+1], lambda x, y: x + y)
tmp = operate_on_Narray(part_neu_vect[0], tmp[a+1], lambda x, y: x / len1)
part_neu_vect = list(tmp)
else:
part_neu_vect = []
2017-12-04 18:01:46 +00:00
############################################ 2
2017-11-28 06:52:47 +00:00
len1 = len(part_neg_vect)
if len1 != 0:
for a in range(len1):
tmp = part_neg_vect[0]
2017-12-04 18:01:46 +00:00
tmp = operate_on_Narray(part_neg_vect[0], tmp[a+1], lambda x, y: x + y)
tmp = operate_on_Narray(part_neg_vect[0], tmp[a+1], lambda x, y: x / len1)
2017-11-28 06:52:47 +00:00
part_neg_vect = list(tmp)
2017-12-04 18:01:46 +00:00
2017-11-28 06:52:47 +00:00
else:
part_neg_vect = []
2017-12-04 18:01:46 +00:00
############################################ 3
2017-11-28 06:52:47 +00:00
len1 = len(part_pos_vect)
if len1 != 0:
for a in range(len1):
tmp = part_pos_vect[0]
tmp = operate_on_Narray(part_pos_vect[0], tmp[a + 1], lambda x, y: x + y)
tmp = operate_on_Narray(part_pos_vect[0], tmp[a + 1], lambda x, y: x / len1)
part_pos_vect = list(tmp)
else:
part_pos_vect = []
2017-12-04 18:01:46 +00:00
############################################ 4
2017-11-28 06:52:47 +00:00
len1 = len(cont_neu_vect)
if len1 != 0:
for a in range(len1):
tmp = cont_neu_vect[0]
tmp = operate_on_Narray(cont_neu_vect[0], tmp[a + 1], lambda x, y: x + y)
tmp = operate_on_Narray(cont_neu_vect[0], tmp[a + 1], lambda x, y: x / len1)
cont_neu_vect = list(tmp)
else:
cont_neu_vect = []
2017-12-04 18:01:46 +00:00
############################################ 5
2017-11-28 06:52:47 +00:00
len1 = len(cont_neg_vect)
if len1 != 0:
for a in range(len1):
tmp = cont_neg_vect[0]
tmp = operate_on_Narray(cont_neg_vect[0], tmp[a + 1], lambda x, y: x + y)
tmp = operate_on_Narray(cont_neg_vect[0], tmp[a + 1], lambda x, y: x / len1)
cont_neg_vect = list(tmp)
else:
cont_neg_vect = []
2017-12-04 18:01:46 +00:00
############################################ 6
2017-11-28 06:52:47 +00:00
len1 = len(cont_pos_vect)
if len1 != 0:
for a in range(len1):
tmp = cont_pos_vect[0]
tmp = operate_on_Narray(cont_pos_vect[0], tmp[a + 1], lambda x, y: x + y)
tmp = operate_on_Narray(cont_pos_vect[0], tmp[a + 1], lambda x, y: x / len1)
cont_pos_vect = list(tmp)
else:
cont_pos_vect = []
2017-12-04 18:01:46 +00:00
############################################ 7
len1 = len(neut_neu_vect)
if len1 != 0:
for a in range(len1):
tmp = neut_neu_vect[0]
tmp = operate_on_Narray(neut_neu_vect[0], tmp[a + 1], lambda x, y: x + y)
tmp = operate_on_Narray(neut_neu_vect[0], tmp[a + 1], lambda x, y: x / len1)
neut_neu_vect = list(tmp)
else:
neut_neu_vect = []
############################################ 8
2017-11-28 06:52:47 +00:00
len1 = len(neut_neg_vect)
if len1 != 0:
for a in range(len1):
tmp = neut_neg_vect[0]
tmp = operate_on_Narray(neut_neg_vect[0], tmp[a + 1], lambda x, y: x + y)
tmp = operate_on_Narray(neut_neg_vect[0], tmp[a + 1], lambda x, y: x / len1)
neut_neg_vect = list(tmp)
else:
neut_neg_vect = []
2017-12-04 18:01:46 +00:00
############################################ 9
2017-11-28 06:52:47 +00:00
len1 = len(neut_pos_vect)
if len1 != 0:
for a in range(len1):
tmp = neut_pos_vect[0]
tmp = operate_on_Narray(neut_pos_vect[0], tmp[a + 1], lambda x, y: x + y)
tmp = operate_on_Narray(neut_pos_vect[0], tmp[a + 1], lambda x, y: x / len1)
neut_pos_vect = list(tmp)
else:
neut_pos_vect = []
2017-11-28 01:14:38 +00:00
2017-11-28 06:52:47 +00:00
return [part_neu_vect, part_neg_vect, part_pos_vect, cont_neu_vect, cont_neg_vect, cont_pos_vect, neut_neu_vect, neut_neg_vect, neut_pos_vect]
2017-11-28 01:14:38 +00:00
2017-12-04 18:01:46 +00:00
def saveTraining():
sert = trainVect()
trnVect = open('trn_vect.vec', 'w')
for i in sert:
trnVect.write(str(i) + '\n')