def post():#登陆主页
f = request.files['file']
input_text = request.form['input_txt'].encode("utf-8-sig")
filename = f.filename
uname = session['username']
if request.form['click'] == '上传文件':
#record user ip , date, user_name
if filename:
curr_time = datetime.datetime.now()
if '.txt' in filename:
content=f.read().decode(encoding="utf-8-sig")
sentences=[sentence for sentence in SentenceSplitter.split(content) if sentence]
# Id=uni_id()
for i in sentences:
sentences_collection.insert({'username':uname,'filename':filename.replace('.txt',''),'sentence':i,'uploadTime':curr_time,'casualtag':'0','eventtag':'0','cflag':'0','eflag':'0'})
return render_template("upload_succeed.html",suc_message='upload successful')
else:
err_message='text format not right!'
return render_template("upload_error.html",err_message=err_message)
else:
curr_time = datetime.datetime.now()
content=input_text
sentences=[sentence for sentence in SentenceSplitter.split(content) if sentence]
for i in sentences:
sentences_collection.insert({'username':uname,'filename':'输入上传','sentence':i,'uploadTime':curr_time,'casualtag':'0','eventtag':'0','cflag':'0','eflag':'0'})
return render_template("upload_succeed.html",suc_message='upload successful')
def collect_infos(word):
infos = hudong.info_extract_hudong(word)
for info in infos:
intro_sents = [
sent for sent in SentenceSplitter.split(info['intro'])
if len(sent) > 0
]
desc_sents = [
sent for sent in SentenceSplitter.split(info['desc'])
if len(sent) > 0
]
print(intro_sents)
print('****' * 5)
print(desc_sents)
def get_sentence_list(self):
"""
获取分句
:return:
"""
sents = SentenceSplitter.split(self.sentence)
return [s for s in sents if s]
def main():
file_list = []
path = r'C:/Users/JeremySun/Desktop/Internship/Project02_corpusProcessor/app'
file_path = batch_file(path=path, file_list=file_list)
for path in file_path:
app = open(path, encoding='utf-8').readlines()
assetPath_text = get_assetPath(text=app)
assetPath_loss_html = loss_html(text=assetPath_text)
assetPath_loss_label = loss_label(text=assetPath_loss_html)
assetPath_loss_mail = loss_mail(text=assetPath_loss_label)
assetPath_loss_other = loss_other(text=assetPath_loss_mail)
assetPath_loss_url = loss_url(text=assetPath_loss_other)
assetPath_clean_url = clean_url(text=assetPath_loss_url)
assetPath_loss_continue = loss_continue(text=assetPath_clean_url)
assetPath_loss_word = loss_word(text=assetPath_loss_continue)
assetPath_loss_comma = loss_comma(text=assetPath_loss_word)
# 分句
assetPath_sentence = SentenceSplitter.split(assetPath_loss_comma)
# 去掉其余符号并写入文件
pattern_all = re.compile(r"[。.；;:：？?!！]") # 加：:
f = open("app_pre.txt", 'a', encoding='utf-8')
for i in tqdm(assetPath_sentence):
i = re.sub(pattern=pattern_all, repl='', string=i)
f.write(i + '\n')
f.close()
def split_sentences(text):
'''
params:
text 文本，包含多句话的文本
划分句子
'''
return SentenceSplitter.split(text)
def sentence_splitter():
"""
分句
"""
sentence = '你好，你觉得这个例子从哪里来的？当然还是直接复制官方文档，然后改了下这里得到的。'
sents = SentenceSplitter.split(sentence) # 分句
print("\n".join(sents))
def get_sentences(self, news):
"""
分句
:param news: str 新闻文本
:return: list 句子列表
"""
return list(SentenceSplitter.split(news))
def __handel(self, news):
news = TextHandle.cht_to_chs(news)
sentences = []
for line in news.strip().split('\n'):
sentences += list(SentenceSplitter.split(line.strip()))
# sentences += [line.strip()]
return sentences
def pad_batch(batch_docs, TEXT):
res_batch_docs = []
max_words, max_sents = 0, 0
res_batch_targets = []
for doc in batch_docs:
doc_text = doc[2]
res_doc_text = []
# 使用LTP将一篇文章划分成若干句子
sents = SentenceSplitter.split(doc_text)
max_sents = max(max_sents, len(sents))
for i, sent in enumerate(sents):
sent = TEXT.preprocess(sent)
sent = [TEXT.vocab.stoi[word] for word in sent]
max_words = max(max_words, len(sent))
res_doc_text.append(sent)
res_batch_docs.append(res_doc_text)
res_batch_targets.append(doc[1])
for doc in res_batch_docs:
sents = doc
for sent in sents:
while len(sent) < max_words:
sent.append(0)
while len(sents) < max_sents:
sents.append([0 for _ in range(max_words)])
return torch.LongTensor(res_batch_docs), torch.LongTensor(
res_batch_targets)
def sentence_split(read_file):
""" 对段落中的句子进行基于符号的划分
:param read_file: 文件txt
:return: 分好的句子存入到sequences了，所以只需要返回状态信息就好了
"""
for paragraph in read_file.readlines():
# 太短的段落（词？）没有分的必要了
if paragraph == '' or len(paragraph) <= 4:
continue
sentence_splitter = SentenceSplitter.split(paragraph)
for sequence in sentence_splitter:
# 去除空行
if sequence == '':
continue
# 二次分隔
second_sentences = re.split('[，,]', sequence)
for second_sentence in second_sentences:
# 对于句子的筛选工作
second_sentence = deal_data(second_sentence)
if second_sentence == '' or len(second_sentence) <= 4:
continue
sentences.add(second_sentence)
str = "分句步骤已完成"
print("=" * 10, str, "=" * 10)
return str
def predict(self, question, text):
input_question = self.sentence2input(question)
sentences = list(SentenceSplitter.split(text))
input_answers_temp = [
self.sentence2input(sentence) for sentence in sentences
]
input_answers_len = np.array([len(a) for a in input_answers_temp])
input_answers = tl.prepro.pad_sequences(input_answers_temp)
input_questions = tl.prepro.pad_sequences([input_question] *
len(sentences))
input_question_len = np.array([len(q) for q in input_questions])
simi_list = self.predict_sess.run(self.simi_P,
feed_dict={
self.Ques: input_questions,
self.Ques_seq_len:
input_question_len,
self.Pos_Ans: input_answers,
self.Pos_Ans_len:
input_answers_len,
self.keep_prob: 1.0
})
# for sen,simi in zip(sentences,simi_list):
# print simi, sen
simi_list = np.array(simi_list)
return sentences[np.argmax(simi_list)], max(simi_list)
def ltpSentenceSplit(content):
"""
ltp split sentence
:param content:
:return:
"""
return list(SentenceSplitter.split(content))
def Relation_Extraction(content):
statements = SentenceSplitter.split(str(content))
relation = json.load(open(resource_path + 'all_relations.json'))
all_triples = []
for statement in statements:
# print(statement)
words, postags, arcs = ltp_parser.parse(statement)
triples = rule_based_extraction(words, postags, arcs)
tri = []
for i in range(len(triples)):
flag = 0
for j in range(len(tri)):
if triples[i][0] == tri[j][0] and triples[i][1] == tri[j][
1] and triples[i][2] == tri[j][2]:
flag = 1
if triples[i][0] == triples[i][2]:
flag = 1
if flag == 0:
tri.append(triples[i])
flag1 = 0
if tri:
# print('关系三元组:')
for triple in tri:
for i in relation:
if (triple[1] == i):
flag1 = 1
break
if (flag1 == 1):
all_triples.append(triple)
return all_triples
def feature_about():
# 获取特征列表
feature_dict = NewsUtil.get_feature()
# 获取新闻中出现特征后最近的5个词及其属性
logger.info("In Prepare Raw News...")
raw_news_data = CommonUtil.read_excel(RAW_NEWS_DEMO_PATH)
raw_news_table = raw_news_data.sheet_by_index(0)
raw_news_rows = raw_news_table.nrows
segmentor = Segmentor() # 初始化实例
segmentor.load_with_lexicon(cws_model_path,
CFETSFX_LEXICON_PATH) # 加载模型，第二个参数是您的外部词典文件路径
feature_about_list = list()
for rowN in range(0, raw_news_rows):
news_content = raw_news_table.cell_value(rowN, 2)
sentences = SentenceSplitter.split(news_content)
for sentence in sentences:
print(sentence)
# 分词
words = segmentor.segment(sentence)
print(list(words))
for word_index in range(0, len(words)):
word = words[word_index]
for feature_word in feature_dict.values():
if feature_word in word:
about_list = list()
count = 0
while word_index < len(words) and count < 6:
about_list.append(words[word_index])
count += 1
word_index += 1
feature_about_list.append(about_list)
print(about_list)
break
segmentor.release()
CommonUtil.write_csv(FEATURE_ABOUT_PATH, feature_about_list)
def main():
file_list = []
path = r'C:\Users\JeremySun\Desktop\Internship\Project02_corpusProcessor\english_text_pre'
file_path = batch_file(path=path, file_list=file_list)
for path in file_path:
english_text_connect = open(path, encoding='utf-8').readlines()
assetPath_loss_url = loss_url(text=english_text_connect)
assetPath_loss_img = loss_img(text=assetPath_loss_url)
assetPath_loss_video = loss_video(text=assetPath_loss_img)
assetPath_loss_src = loss_src(text=assetPath_loss_video)
assetPath_loss_div = loss_div(text=assetPath_loss_src)
assetPath_loss_span = loss_span(text=assetPath_loss_div)
assetPath_loss_p = loss_p(text=assetPath_loss_span)
assetPath_loss_special = loss_special(text=assetPath_loss_p)
assetPath_loss_continue = loss_continue(text=assetPath_loss_special)
assetPath_loss_word = loss_word(text=assetPath_loss_continue)
assetPath_loss_chino = loss_chino(text=assetPath_loss_word)
assetPath_loss_greek = loss_greek(text=assetPath_loss_chino)
assetPath_loss_pinyin = loss_pinyin(text=assetPath_loss_greek)
assetPath_loss_fake = loss_fake(text=assetPath_loss_pinyin)
assetPath_loss_tradition = loss_tradition(text=assetPath_loss_fake)
assetPath_loss_comma = loss_comma(text=assetPath_loss_tradition)
# 分句
english_text_sentence = SentenceSplitter.split(assetPath_loss_comma)
# 去掉其余符号并写入文件
pattern_all = re.compile(r"[。.；;？?!！]")
f = open("english_text_sentence_pre.txt", 'a', encoding='utf-8')
for i in tqdm(english_text_sentence):
i = re.sub(pattern=pattern_all, repl='', string=i)
f.write(i + '\n')
f.close()
def split2sent(self, text):
'''
对文本进行分句
'''
from pyltp import SentenceSplitter
sents = SentenceSplitter.split(text)
return sents
def article_preprocessing(article):
"""
对News进行预处理，对文本进行分词和词性标注
这里的分词,没有去掉stopwords,也没有去掉标点符号
Args:
Article: 原始的新闻Article
Return：
预处理之后的新闻PreprocessArticle
"""
# print("article preprocessing")
pre_article = PreprocessArticle()
# pre_article.id = article.id
pre_article.date = article.time
pre_article.title = article.title
pre_article.seg_pos_title = seg_pos.cut(article.title)
sentences = SentenceSplitter.split(article.content) # Using ltp split sentence
sentences = [x for x in sentences if x != '']
for idx, x in enumerate(sentences):
sentence = Sentence()
sentence.text = x
sentence.location = idx
sentence.seg_pos = seg_pos.cut(x)
pre_article.sentences.append(sentence)
return pre_article
def tokenized_sub_sents(row):
content = row.iloc[1]
sub_sents = []
sub_tags = []
sents = SentenceSplitter.split(content)
for sent in sents:
subs = [x for x in re.split(punc_pattern, sent) if x]
subss = [
jieba.posseg.cut(x, HMM=False) for x in subs
if not re.findall(no_chn, x)
]
tags = []
subs = []
for s in subss:
tag = []
sub = []
for t0, t1 in s:
tag.append(t1)
sub.append(t0)
tags.append(tag)
subs.append(sub)
assert len(tags) == len(subs)
sub_sents.extend(subs)
sub_tags.extend(tags)
# print(sub_sents, sub_tags)
row["sub_sents_tokenized"] = sub_sents
row["sub_sents_postagged"] = sub_tags
return row
def mainLocaltion(self, dirName="西游记白话文"):
txtlist = os.listdir(self.book_root_path + dirName)
lo_list_book = []
for txt in txtlist:
lo_list_chapter = []
print txt
lines = self.readBookLines(self.book_root_path + dirName + "/" +
txt)
for line in lines:
if line != "":
sents = SentenceSplitter.split(line)
for sent in sents:
words_line = self.segmentor(sent)
lo_list_line = self.posttagerNLNS(words_line)
lo_list_chapter += lo_list_line
# 统计每一章节top 10
top_itf_chapter, top_lo_chapter = self.getTopTen(lo_list_chapter)
lo_list_book += top_lo_chapter
self.writeTxt(self.mainlo_root_path + dirName + "/" + txt,
top_itf_chapter)
print txt + "本章节top 10----------------------"
for cloname, clotimes, clofreq in top_itf_chapter:
print cloname, clotimes, clofreq
# 统计整本书 top 10
top_loitf_book, top_lo_book = self.getTopTen(lo_list_book)
self.writeTxt(self.mainlo_root_path + dirName + "/AllChapter.txt",
top_loitf_book)
print "整本书 top 10----------------------"
for bloname, blotimes, blofreq in top_loitf_book:
print bloname, blotimes, blofreq
def mainName(self, dirName):
txtlist = os.listdir(self.book_root_path + dirName)
name_list_book = []
for txt in txtlist:
name_list_chapter = []
print txt
lines = self.readBookLines(self.book_root_path + dirName + "/" +
txt)
for line in lines:
if line != "":
sents = SentenceSplitter.split(line)
for sent in sents:
words_line = self.segmentor(sent)
postags_line, name_list_line = self.posttaggerNH(
words_line)
name_list_chapter += name_list_line
# 统计每一章节top 10
top_itf_chapter, top_name_chapter = self.getTopTen(
name_list_chapter) # [(name,times,freq),()]
name_list_book += top_name_chapter
self.writeTxt(self.mainrole_root_path + dirName + "/" + txt,
top_itf_chapter)
print txt + "本章节top 10----------------------"
for cname, ctimes, cfreq in top_itf_chapter:
print cname, ctimes, cfreq
# 统计整本书 top 10
top_itf_book, top_name_book = self.getTopTen(name_list_book)
self.writeTxt(self.mainrole_root_path + dirName + "/AllChapter.txt",
top_itf_book)
print "整本书 top 10----------------------"
for bname, btimes, bfreq in top_itf_book:
print bname, btimes, bfreq
def sentence_split(paras_aft_retri, topic_to_search):
assert len(paras_aft_retri) == len(topic_to_search)
res = []
for i, item in enumerate(paras_aft_retri):
para_to_split, urls, tops, query = item
tops = topic_to_search[i][1]
sent_aft_split = []
url_aft = []
for ind, para in enumerate(para_to_split):
sents = SentenceSplitter.split(para)
url = urls[ind]
for sent in sents:
if sent[-1] == '？' or sent[-1] == '?':
continue
elif "引用" in sent or "相关阅读" in sent or "特别声明" in sent or "声明" in sent or "来源" in sent \
or "原标题" in sent or "联系电话" in sent or "联系方式" in sent:
continue
elif len(sent) < 15:
continue
elif check_contain_chinese(sent) < 0.6:
continue
else:
sent_aft_split.append(sent)
url_aft.append(url)
assert len(sent_aft_split) == len(url_aft)
res.append((sent_aft_split, url_aft, tops, query))
# print(res)
return res
def main():
file_list = []
path = r'D:\实习数据备份\备份\professional'
file_path = batch_file(path=path, file_list=file_list)
for path in file_path:
profession_text_connect = open(path, encoding='utf-8').readlines()
assetPath_loss_html = loss_html(text=profession_text_connect)
assetPath_loss_label = loss_label(text=assetPath_loss_html)
assetPath_loss_mail = loss_mail(text=assetPath_loss_label)
assetPath_loss_other = loss_other(text=assetPath_loss_mail)
assetPath_loss_url = loss_url(text=assetPath_loss_other)
assetPath_clean_url = clean_url(text=assetPath_loss_url)
assetPath_loss_continue = loss_continue(text=assetPath_clean_url)
assetPath_loss_word = loss_word(text=assetPath_loss_continue)
assetPath_loss_comma = loss_comma(text=assetPath_loss_word)
# 分句
assetPath_sentence = SentenceSplitter.split(assetPath_loss_comma)
# 去掉其余符号并写入文件
pattern_all = re.compile(r"[。.；;:：？?!！/|]") # 加：:
f = open("profession_pre.txt", 'a', encoding='utf-8')
for i in tqdm(assetPath_sentence):
i = re.sub(pattern=pattern_all, repl='', string=i)
f.write(i + '\n')
f.close()
def main():
file_list = []
path = r'C:\Users\JeremySun\Desktop\Internship\Project02_corpusProcessor\english_folder'
file_path = batch_file(path=path, file_list=file_list)
for path in file_path:
english_text_connect = open(path, encoding='utf-8').readlines()
assetPath_loss_html = loss_html(text=english_text_connect)
assetPath_loss_label = loss_label(text=assetPath_loss_html)
assetPath_loss_mail = loss_mail(text=assetPath_loss_label)
assetPath_loss_other = loss_other(text=assetPath_loss_mail)
assetPath_loss_url = loss_url(text=assetPath_loss_other)
assetPath_clean_url = clean_url(text=assetPath_loss_url)
assetPath_loss_continue = loss_continue(text=assetPath_clean_url)
assetPath_loss_word = loss_word(text=assetPath_loss_continue)
assetPath_loss_comma = loss_comma(text=assetPath_loss_word)
# 分句
english_text_sentence = SentenceSplitter.split(assetPath_loss_comma)
# 去掉其余符号并写入文件
pattern_all = re.compile(r"[。.；;？?!！:：]") # 加：:
pattern_last = re.compile(r'[a-zA-Z0-9]{13,}')
f = open("english_text_sent_pre.txt", 'a', encoding='utf-8')
for i in tqdm(english_text_sentence):
if len(i) <= 100:
i = re.sub(pattern=pattern_all, repl=' ', string=i)
i = re.sub(pattern=pattern_last, repl='', string=i)
f.write(i.strip() + '\n')
f.close()
def get_roles_emotions(self):
'''
获取self.book中的人物及情绪变化指标
:return:
'''
contents = self.get_chapterList()
roles = self.get_roles()
role_points = [[] for i in range(len(roles))]
for content in contents:
sents = SentenceSplitter.split(content)
sent_words = []
for sent in sents:
words = list(self.segmentor.segment(sent))
for role in roles:
if role in words:
sent_words.append(words)
break
role_emotionWords = self.get_role_emotionWord(roles, sent_words)
points = self.count_point(role_emotionWords, roles)
for i in range(len(role_points)):
role_points[i].append(points[i])
for term in range(role_points.__len__()):
index = role_points[term].__len__() - 2
while index >= 1:
if role_points[term][index] == 0 or role_points[term][
index] == 0.0:
role_points[term][index] = (
role_points[term][index - 1] +
role_points[term][index + 1]) / 2
index -= 1
for i in range(len(role_points)):
print roles[i], role_points[i]
return roles, role_points
def getDAN(self,path = "测试测试啊哈哈哈"):
'''
读取文本文件,获取danlist
:return:
'''
#txts = []
dan_list = []
files = os.listdir(self.book_root_path)
for file in files:
fileposition = self.book_root_path + "\\" + file
print("file name:", fileposition)
with open(fileposition,"r",encoding="utf-8")as f:
lines = f.readlines()
for line in lines :
#txts.append(line)
if line != "":
sentences = SentenceSplitter.split(line)
#print("sentences:", sentences)
for sentence in sentences:
words = self.segmentor(sentence)
#print("sentences 11:",words)
dan_list_line = self.postagger(words)
dan_list += dan_list_line
f.close()
return list(set(dan_list))
#def getDAN(self):
'''
def main():
file_list = []
path = r'D:\实习数据备份\备份\xinwenshuju'
file_path = batch_file(path=path, file_list=file_list)
for path in file_path:
try:
news_text_connect = open(path, encoding='utf-8').readlines()
assetPath_loss_html = loss_html(text=news_text_connect)
assetPath_loss_label = loss_label(text=assetPath_loss_html)
assetPath_loss_mail = loss_mail(text=assetPath_loss_label)
assetPath_loss_other = loss_other(text=assetPath_loss_mail)
assetPath_loss_url = loss_url(text=assetPath_loss_other)
assetPath_clean_url = clean_url(text=assetPath_loss_url)
assetPath_loss_continue = loss_continue(text=assetPath_clean_url)
assetPath_loss_word = loss_word(text=assetPath_loss_continue)
assetPath_loss_comma = loss_comma(text=assetPath_loss_word)
# 分句
assetPath_sentence = SentenceSplitter.split(assetPath_loss_comma)
# 去掉其余符号并写入文件
pattern_all = re.compile(r"[。.；;:：？?!！|]") # 加：:
f = open("news_pre.txt", 'a', encoding='utf-8')
for i in tqdm(assetPath_sentence):
i = re.sub(pattern=pattern_all, repl='', string=i)
f.write(i + '\n')
f.close()
except:
print("utf-8 codec can not decode byte 0xc3 in position 0")
def get_sentences(content_sentences):
sentences_list = list(SentenceSplitter.split(content_sentences))
# delete empty member in list
sentences_list = [item.strip() for item in sentences_list if item]
return sentences_list
def sentence_split(self, string):
"""
分句
:param string: 传入字符串
:return:
"""
sentence = SentenceSplitter.split(string)
return sentence
def split_sentence(sentence):
# pattern = re.compile('[。；]')
# split = pattern.sub('\r', sentence).split() # split sentence
# return split
sents = SentenceSplitter.split(sentence)
result = [x for x in sents if x != '']
return result
def sentence_splitter(sentence):
sents = SentenceSplitter.split(sentence) # 分句
sent=''
for i in sents:
i=re.sub("[\s+\.\!\/_,$%^*(+\"\')]+|[+——()?【】“”！，。？、~@#￥%……&*（）]+", "",str(i))
sent+=i+' '
#print(sent)
return sent
#!/usr/bin/env python # -*- coding: utf-8 -*- import sys, os ROOTDIR = os.path.join(os.path.dirname(__file__), os.pardir) sys.path = [os.path.join(ROOTDIR, "lib")] + sys.path # Set your own model path MODELDIR=os.path.join(ROOTDIR, "ltp_data") from pyltp import SentenceSplitter, Segmentor, Postagger, Parser, NamedEntityRecognizer, SementicRoleLabeller paragraph = '中国进出口银行与中国银行加强合作。中国进出口银行与中国银行加强合作！' sentence = SentenceSplitter.split(paragraph)[0] segmentor = Segmentor() segmentor.load(os.path.join(MODELDIR, "cws.model")) words = segmentor.segment(sentence) print "\t".join(words) postagger = Postagger() postagger.load(os.path.join(MODELDIR, "pos.model")) postags = postagger.postag(words) # list-of-string parameter is support in 0.1.5 # postags = postagger.postag(["中国","进出口","银行","与","中国银行","加强","合作"]) print "\t".join(postags) parser = Parser() parser.load(os.path.join(MODELDIR, "parser.model")) arcs = parser.parse(words, postags)