def sohu_cmt(request):
if off:
return HttpResponse(
"<h1>网站暂停使用, 项目 <a href=\"http://tofind.space/NCspider/\">Github 主页</a> </h1>"
)
try:
#return HttpResponse(str(sohu_comment.objects.filter(news_id=comment_id).values("author")))#留作DEBUG用
comment_id = request.GET.get('comment_id')
news_title = request.GET.get('title')
lastern_comments = sohu_comment.objects.filter(news_id=comment_id)[:50]
attitude_tendency = []
for com in lastern_comments:
s = nlp(com.comment)
senti_number = s.sentiments
attitude_tendency.append(senti_number)
rgb_list = ["#565D8E", "#499AA1", "#FFFFFF", "#F9E063", "#BBDC37"]
rgb_level = 2
if senti_number > 0.7 and senti_number < 0.8:
rgb_level = 3
elif senti_number >= 0.8:
rgb_level = 4
elif senti_number < 0.35 and senti_number > 0.25:
rgb_level = 1
elif senti_number <= 0.25:
rgb_level = 0
#senti_rgb = str(hex(int( senti_number * 0xFFFFFF))).replace("0x", "#").upper()
senti_rgb = rgb_list[rgb_level]
com.sense_color = senti_rgb
if len(attitude_tendency) > 0:
tendency_number = sum(attitude_tendency) / len(attitude_tendency)
attitude_list = ["很消极", "消极", "中立", "积极", "很积极"]
attitude = attitude_list[int(round(tendency_number * 10 / 2)) - 1]
else:
attitude = "不得而知"
return render(
request, 'news_opin/sohu_comment_list.html', {
'sohu_comments_list': lastern_comments,
'attitude': attitude + "?",
'news_title': news_title,
})
except Exception as e:
return render(request, 'news_opin/error_info.html', {
'error_info': str(e),
})
def sohu_cmt(request):
if off:
return HttpResponse('<h1>网站暂停使用, 项目 <a href="http://tofind.space/NCspider/">Github 主页</a> </h1>')
try:
# return HttpResponse(str(sohu_comment.objects.filter(news_id=comment_id).values("author")))#留作DEBUG用
comment_id = request.GET.get("comment_id")
news_title = request.GET.get("title")
lastern_comments = sohu_comment.objects.filter(news_id=comment_id)[:50]
attitude_tendency = []
for com in lastern_comments:
s = nlp(com.comment)
senti_number = s.sentiments
attitude_tendency.append(senti_number)
rgb_list = ["#565D8E", "#499AA1", "#FFFFFF", "#F9E063", "#BBDC37"]
rgb_level = 2
if senti_number > 0.7 and senti_number < 0.8:
rgb_level = 3
elif senti_number >= 0.8:
rgb_level = 4
elif senti_number < 0.35 and senti_number > 0.25:
rgb_level = 1
elif senti_number <= 0.25:
rgb_level = 0
# senti_rgb = str(hex(int( senti_number * 0xFFFFFF))).replace("0x", "#").upper()
senti_rgb = rgb_list[rgb_level]
com.sense_color = senti_rgb
if len(attitude_tendency) > 0:
tendency_number = sum(attitude_tendency) / len(attitude_tendency)
attitude_list = ["很消极", "消极", "中立", "积极", "很积极"]
attitude = attitude_list[int(round(tendency_number * 10 / 2)) - 1]
else:
attitude = "不得而知"
return render(
request,
"news_opin/sohu_comment_list.html",
{"sohu_comments_list": lastern_comments, "attitude": attitude + "?", "news_title": news_title},
)
except Exception as e:
return render(request, "news_opin/error_info.html", {"error_info": str(e)})
import pandas as pd
from snownlp import SnowNLP as nlp
filePath = '../corpus/moods/喜悦.csv'
df = pd.read_csv(filePath).sample(1000)
count = 0
for index, row in df.iterrows():
content = row['review']
s = nlp(content)
# print(s.sentiments)
if s.sentiments < 0.7:
print(s.sentiments, content)
count += 1
print(count/df.shape[0])
# 利用snownlp交叉印证数据的质量，喜悦的准确度只有大约70%
def _snow(self):
return nlp(self.text)
def sent_senti(self):
# filtered = filter(sentence_contains_brace, self._snow.sentences)
sentences = self._snow.sentences
return {_: nlp(_).sentiments for _ in sentences}
def checkDataNLP(self, factorDF=None, code_=None):
"""
判断TXT的情感偏向
:param self:
:param txt: df
:return: 返回TXT的情感偏向sentments
"""
content_title = self.contentReplace(factorDF['title']) # 删除空行/多余的空格
content_datas = self.contentReplace(factorDF['datas']) # 删除空行/多余的空格
# # title的情感色彩分析
factorDF['title_Marking'] = map(
lambda x: 0 if x == '0' or len(x) == 0 else
(float(nlp(x).sentiments)), content_title)
# # datas的情感色彩分析
factorDF['datas_Marking'] = map(
lambda x: 0 if x == '0' or len(x) == 0 else
(float(nlp(x).sentiments)), content_datas)
upTimeValues = factorDF['upTime'].values
factorDF['upTime'] = map(self.filterDateTime, upTimeValues)
#print "factorDF['upTime']==",factorDF['upTime']
factorDFNew = factorDF[[
'code', 'title_Marking', 'datas_Marking', 'upTime'
]]
upTimeList = np.array(factorDFNew['upTime'].values)
indexListNew = [
'0' if len(n.strip()) < 10 else (n.strip()) for n in upTimeList
]
factorDFNew['upTime'] = indexListNew
# 开始groupby 汇总每一天的数据
factorData_DF = pd.DataFrame(
data=factorDFNew.values,
columns=["code", "titleMarketing", "datasMarketing", "dateUpTime"])
factorData_DF['dateUpTime'] = map(self.secondFilterDate,
factorData_DF['dateUpTime'])
factorDFLast = factorData_DF[factorData_DF['dateUpTime'] != '0']
code_list = factorDFLast['code'].values
factorDFLast.drop('code', axis=1, inplace=True)
resultDF = pd.DataFrame(
index=factorDFLast['dateUpTime'].values,
data=factorDFLast.values,
columns=['titleMarketing', 'datasMarketing', 'dateUpTime'])
resultDF['titleMarketing'] = map(lambda x: float(x),
resultDF['titleMarketing'].values)
resultDF['datasMarketing'] = map(lambda x: float(x),
resultDF['datasMarketing'].values)
resultDF.sort_values(by=['dateUpTime'], inplace=True)
new_data = resultDF.groupby('dateUpTime').mean()
newCount = len(new_data)
new_data['code'] = code_list[:newCount]
new_data['dateUpTime'] = new_data.index
# 根据历史交易日，填充所有缺失数据
lostDate = [] # 把所有缺失时间收集起来一起填充
index_data = new_data.index
for dt in self.indexDateList:
if len(dt) < 10: pass
else:
dt = dt + " 00:00:00"
if dt in index_data: continue
else: lostDate.append(dt)
for dt in lostDate:
new_data.loc[dt] = {
'titleMarketing': 0,
'datasMarketing': 0,
'code': code_,
'dateUpTime': dt
}
new_data.sort_values(by=['dateUpTime'], inplace=True)
new_data.set_index('dateUpTime')
new_data['titleMarketing'] = map(
lambda x: x if x > .5 else (0 if x == 0 else (x - .5)),
new_data['titleMarketing'].values)
new_data['datasMarketing'] = map(
lambda x: x if x > .5 else (0 if x == 0 else (x - .5)),
new_data['datasMarketing'].values)
new_data['title_Rolling5_mean'] = np.array(
new_data['titleMarketing'].rolling(min_periods=1,
window=5).mean().values)
new_data['title_Rolling10_mean'] = np.array(
new_data['titleMarketing'].rolling(min_periods=1,
window=10).mean().values)
new_data['title_Rolling20_mean'] = np.array(
new_data['titleMarketing'].rolling(min_periods=1,
window=20).mean().values)
new_data['title_Rolling30_mean'] = np.array(
new_data['titleMarketing'].rolling(min_periods=1,
window=30).mean().values)
new_data['datas_Rolling5_mean'] = np.array(
new_data['datasMarketing'].rolling(min_periods=1,
window=5).mean().values)
new_data['datas_Rolling10_mean'] = np.array(
new_data['datasMarketing'].rolling(min_periods=1,
window=10).mean().values)
new_data['datas_Rolling20_mean'] = np.array(
new_data['datasMarketing'].rolling(min_periods=1,
window=20).mean().values)
new_data['datas_Rolling30_mean'] = np.array(
new_data['datasMarketing'].rolling(min_periods=1,
window=30).mean().values)
#date__=str(datetime.datetime.now()).replace(':','').replace('.','')
#df.to_csv(r'{}_{}.csv'.format(columnName_,date__),encoding='utf_8_sig')
#print u'写入完成'
return new_data
from snownlp import SnowNLP as nlp s1 = nlp(u'hahaha') s2 = nlp(u'哈哈哈') s3 = nlp(u'哈哈哈哈哈哈') s4 = nlp(u'哈哈哈哈哈哈哈哈哈') print(s1.sentiments, s2.sentiments, s3.sentiments, s4.sentiments) s1 = nlp(u'开心') s2 = nlp(u'快乐') s3 = nlp(u'不开心') s4 = nlp(u'不快乐') print(s1.sentiments, s2.sentiments, s3.sentiments, s4.sentiments) s1 = nlp(u'有趣') s2 = nlp(u'无聊') s3 = nlp(u'傻逼') s4 = nlp(u'天使') print(s1.sentiments, s2.sentiments, s3.sentiments, s4.sentiments) s1 = nlp(u'不知所云') s2 = nlp(u'耗子尾汁') s3 = nlp(u'不明觉厉') s4 = nlp(u'阴阳怪气') print(s1.sentiments, s2.sentiments, s3.sentiments, s4.sentiments) s1 = nlp(u'开心')