-
Notifications
You must be signed in to change notification settings - Fork 13
/
raw_data_clean.py
185 lines (172 loc) · 7.46 KB
/
raw_data_clean.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
# -*- coding: utf-8 -*-
from hanziconv import HanziConv
from joblib import Parallel, delayed
import pandas as pd
from collections import Counter
import re
import jieba
import random
def load_user_dict(dictfile='data/dict.csv'):
dict_words=pd.read_csv(dictfile)
dict_words=dict_words[dict_words.freq>9]
n=dict_words.shape[0]
for i in range(0,n):
try:
jieba.add_word(dict_words.char[i],int(dict_words.len[i])*1000,'nz')
except Exception as e:
print(e)
stopwords=open('data/stopwords.txt').readlines()
stopwords=[w.strip() for w in stopwords]
def words_seg(x):
w = jieba.cut(x.upper())
w = [ word.strip() for word in w if word not in stopwords]
w=[word.strip() for word in w if len(word)>0]
return [' '.join(w), len(w)]
def seg(k):
words,words_size=words_seg(X.content[k])
return X.shop_url[k],X.post_time[k],X.content[k],X.score[k],words,words_size
X=pd.read_csv('data/comments_raw_v3.csv')
'''
shop_url=X.shop_url.tolist()
deal_url=X.deal_url.tolist()
mn=len(shop_url)
shop=[]
for i in range(0,mn):
if len(str(shop_url[i]))<6:
shop.append(deal_url[i])
else:
shop.append(shop_url[i])
X=X.loc[:,['post_time', 'content', 'score']]
X.insert(0,'shop_url',shop)
'''
def to_S(k):
txt=X.content[k].strip()
txt = re.sub('\t|\r', '\n',txt)
txt = txt.replace('\n\n', '\n')
txt = re.sub(' |\u3000', ' ', txt)
txt=HanziConv.toSimplified(txt)
txt=txt.strip()
return (X.shop_url[k], X.post_time[k],txt,int(X.score[k]),len(txt))
n=X.shape[0]
'''
X1= Parallel(n_jobs=18, verbose=1, pre_dispatch='2*n_jobs')(delayed(to_S)(k) for k in range(0,n))
X1=[x[:4] for x in X1 if x[4]>0]
X1=pd.DataFrame(X1,columns=['shop_url', 'post_time', 'content', 'score'])
X=X1.sort_values(by="score")
'''
load_user_dict(dictfile='data/dict_utf8.csv')
n=X.shape[0]
X2= Parallel(n_jobs=16, verbose=1, pre_dispatch='2*n_jobs')(delayed(seg)(k) for k in range(0,n))
X2=[x for x in X2 if x[5]>0]
X2=pd.DataFrame(X2,columns=['shop_url', 'post_time', 'content', 'score', 'words','words_size'])
X2=X2.sort_values(by="score")
ids=[i+1 for i in range(0,X2.shape[0])]
X2.insert(0,'id',ids)
X2.to_csv('data/comments_words_v4.csv',index=False,encoding="UTF-8")
X3=X2.loc[:,['id','score','words','words_size']]
X3.to_csv('data/comments_words_std_v4.csv',index=False,encoding="UTF-8")
X4=X3[X3.words_size>11]
X4=X4[X4.words_size<141]
Counter(X4.score.tolist())
Counter(X4.words_size.tolist())
X4.to_csv('data/comments_words_clean_std_v4.csv',index=False,encoding="UTF-8")
#### 句子切分，作为训练词向量语料
sentences= '\n'.join(X.content.tolist())
sentences=re.split('[\n\t\r。]',sentences)
sentences=[s for s in sentences if len(s)>0]
sentences=list(set(sentences))
sentences_words= Parallel(n_jobs=20, verbose=1, pre_dispatch='2*n_jobs')(delayed(words_seg)(k) for k in sentences)
sentences_words=[x[0] for x in sentences_words if x[1]>0]
sentences_words1=[x+'\n' for x in sentences_words ]
fo = open("data/comments_w2v_train_words_v4.txt", "w")
fo.writelines( sentences_words1 )
fo.close()
#################################################################
def gen_sample(datafile='data/comments_words_v3.csv',
simfile='data/comments_words_std_v3_sim.csv',
outfile='data/comments_v3'):
A1=pd.read_csv(datafile)
#words_char=A1.words.tolist()
#words_char_num=[len(x.replace(' ','')) for x in words_char]
#A1.insert(7, 'words_char_size', words_char_num)
A2=pd.read_csv(simfile)
A3= A1[A1.id.isin(A2.id) ]
A4=pd.merge(A3,A2,on='id',how='left')
A4=A4.drop('score_y',axis=1)
A4.columns=['id', 'shop_url', 'post_time', 'content', 'score', 'words', 'words_size', 'sim_id', 'sim_score', 'sim_len']
#A4.columns = ['id', 'score', 'words', 'words_size', 'sim_id', 'sim_score', 'sim_len']
#A4=A4[A4.words_size>15]
print('data size:',A4.shape[0])
A4[A4.sim_len>1].to_csv(outfile+'_sim_view.csv',index=False,encoding='UTF-8')
#A5 = A4[A4.id == A4.sim_id]
#print('del sim data :', A5.shape[0])
A6=A4[A4.sim_len==1]
print('no sim data :', A6.shape[0])
A6=A6.loc[:,['id','score','words']]
scores_stat=Counter(A6.score.tolist())
print('score dis:',scores_stat)
scores_stat=[(s1,s2) for s1,s2 in scores_stat.items()]
scores_stat=sorted(scores_stat, key=lambda x: x[1],reverse=False)
n_sample=scores_stat[0][1]
#random.sample(range(0,),n_sample)
Z1=A6[A6.score==1]
#Z1=Z1.iloc[sorted(random.sample(range(0,Z1.shape[0]),n_sample))]
Z1=Z1.sample(frac=n_sample/Z1.shape[0])
Z2=A6[A6.score==2]
Z2=Z2.sample(frac=n_sample/Z2.shape[0])
Z3=A6[A6.score==3]
Z3=Z3.sample(frac=n_sample/Z3.shape[0])
Z4=A6[A6.score==4]
Z4=Z4.sample(frac=n_sample/Z4.shape[0])
Z5=A6[A6.score==5]
Z5=Z5.sample(frac=n_sample/Z5.shape[0])
Z=pd.concat([Z1,Z2,Z3,Z4,Z5])
Z=Z.sample(frac=1).reset_index(drop=True)
print('sample data :', Z.shape[0])
Z.to_csv(outfile+'_words_sample.csv',index=False,encoding='UTF-8')
#Z_train=Z.sample(frac=0.8).reset_index(drop=True)
#Z_train.to_csv(outfile+'_train_sample.csv',index=False,encoding='UTF-8')
#Z_test=Z[~Z.id.isin(Z_train.id)]
#Z_test.to_csv(outfile+'_test_sample.csv',index=False,encoding='UTF-8')
gen_sample(datafile='data/comments_words_v5_L4_7.csv',simfile='data/comments_words_sim_v3_32_0.85_L4-7.csv',outfile='data/comments_v5_32_0.60_L4-7')
gen_sample(datafile='data/comments_words_v5.csv',simfile='data/comments_words_std_v4_64_0.60_L4-10.csv',outfile='data/comments_v4_64_0.60_L4-10')
A1=pd.read_csv('data/comments_words_v4.csv')
s=Counter(A1.words)
s1=[k for k,v in s.items() if v==1]
A2= A1[A1.words.isin(s1) ]
A2.to_csv('data/comments_words_v5.csv',index=False,encoding="UTF-8")
A3=A2.loc[:,['id','score','words','words_size']]
n1=51;n2=120
A31=A3[A3.words_size>=n1]
A32=A31[A31.words_size<=n2]
A32.to_csv('data/comments_words_v5_L'+str(n1)+'_'+str(n2)+'.csv',index=False,encoding="UTF-8")
'''
content=X1.content.tolist()
char=list(' '.join(content))
fq=Counter(char)
freq=[(x[0],x[1]) for x in fq.items()]
freq=pd.DataFrame(freq,columns=['char','freq'])
freq=freq.sort_values(by="freq",ascending=False)
freq.to_csv('data/freq.csv',index=False)
'''
# sp="[，。！#….～\\?,【】!\\^\\)\\(；）（-_：⊙”“/\\\\\[\\]\\+￣′·`;@=．\\|>‘<°｀」》《’丶﹏˙￥&ノ＿\\$ \n\t]"
'''
## 统计常用词语，作为扩展词库
def replace_char(x):
sp="[˵⁻♡♀_◎ω＼↗↖｢」》《•๑⊙′/╯▽╰～ಡ·•●з」∠ε`~。，；：“”‘’？【】#￥（）、！,!':;\\|\\-\\*\\\…\\.\\[\\]\\(\\)\\?\n\r\t ]|[\uD800-\uDBFF][\uDC00-\uDFFF]|[0-9A-Za-z]|\u3000"
x=re.sub(sp,' ',x)
return x.strip()
txt= Parallel(n_jobs=18, verbose=1, pre_dispatch='2*n_jobs')(delayed(replace_char)(k) for k in X1.content)
txt=' '.join(txt)
txt=txt.split(' ')
sentences=Counter(txt)
sentences=[(x[0],x[1],len(x[0])) for x in sentences.items()]
sentences=pd.DataFrame(sentences,columns=['char','freq','len'])
sentences=sentences.sort_values(by="freq",ascending=False)
sentences_1=sentences[sentences.freq>50 ]
sentences_1=sentences_1[ sentences_1.len<6]
sentences_1=sentences_1[ sentences_1.len>1]
sentences_1.to_csv('data/sentences.csv',index=False,encoding="GBK")
'''
#import thulac
#thu1 = thulac.thulac(user_dict="data/thu_dict.txt", model_path=None, T2S=True, seg_only=True, filt=True, deli='___')