今日头条
阅读目录
3 自然语言处理手记
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148 | #!/usr/bin/env python# -*- coding:utf-8 -*-import jiebaimport osimport reimport timefrom jpype import * '''title:利用结巴分词进行文本语料的批量处理 1 首先对文本进行遍历查找 2 创建原始文本的保存结构 3 对原文本进行结巴分词和停用词处理 4 对预处理结果进行标准化格式,并保存原文件结构路径author:白宁超myblog:http://www.cnblogs.com/baiboy/time:2017年4月28日10:03:09''' '''创建文件目录path:根目录下创建子目录'''def mkdir(path): # 判断路径是否存在 isExists=os.path.exists(path) # 判断结果 if not isExists: os.makedirs(path) print(path+' 创建成功') return True else: pass print('-->请稍后,文本正在预处理中...') '''结巴分词工具进行中文分词处理:read_folder_path:待处理的原始语料根路径write_folder_path 中文分词经数据清洗后的语料'''def CHSegment(read_folder_path,write_folder_path): stopwords ={}.fromkeys([line.strip() for line in open('../Database/stopwords/CH_stopWords.txt','r',encoding='utf-8')]) # 停用词表 # 获取待处理根目录下的所有类别 folder_list = os.listdir(read_folder_path) # 类间循环 # print(folder_list) for folder in folder_list: #某类下的路径 new_folder_path = os.path.join(read_folder_path, folder) # 创建一致的保存文件路径 mkdir(write_folder_path+folder) #某类下的保存路径 save_folder_path = os.path.join(write_folder_path, folder) #某类下的全部文件集 # 类内循环 files = os.listdir(new_folder_path) j = 1 for file in files: if j > len(files): break # 读取原始语料 raw = open(os.path.join(new_folder_path, file),'r',encoding='utf-8').read() # 只保留汉字 # raw1 = re.sub("[A-Za-z0-9\[\`\~\!\@\#\$\^\&\*\(\)\=\|\{\}\'\:\;\'\,\[\]\.\<\>\/\?\~\!\@\#\\\&\*\%]", "", raw) # jieba分词 wordslist = jieba.cut(raw, cut_all=False) # 精确模式 # 停用词处理 cutwordlist='' for word in wordslist: if word not in stopwords and word=="\n": cutwordlist+="\n" # 保持原有文本换行格式 elif len(word)>1 : cutwordlist+=word+"/" #去除空格 #保存清洗后的数据 with open(os.path.join(save_folder_path,file),'w',encoding='utf-8') as f: f.write(cutwordlist) j += 1 '''结巴分词工具进行中文分词处理:read_folder_path:待处理的原始语料根路径write_folder_path 中文分词经数据清洗后的语料'''def HanLPSeg(read_folder_path,write_folder_path): startJVM(getDefaultJVMPath(), "-Djava.class.path=C:\hanlp\hanlp-1.3.2.jar;C:\hanlp", "-Xms1g", "-Xmx1g") # 启动JVM,Linux需替换分号;为冒号: stopwords ={}.fromkeys([line.strip() for line in open('../Database/stopwords/CH_stopWords.txt','r',encoding='utf-8')]) # 停用词表 # 获取待处理根目录下的所有类别 folder_list = os.listdir(read_folder_path) # 类间循环 # print(folder_list) for folder in folder_list: #某类下的路径 new_folder_path = os.path.join(read_folder_path, folder) # 创建一致的保存文件路径 mkdir(write_folder_path+folder) #某类下的保存路径 save_folder_path = os.path.join(write_folder_path, folder) #某类下的全部文件集 # 类内循环 files = os.listdir(new_folder_path) j = 1 for file in files: if j > len(files): break # 读取原始语料 raw = open(os.path.join(new_folder_path, file),'r',encoding='utf-8').read() # HanLP分词 HanLP = JClass('com.hankcs.hanlp.HanLP') wordslist = HanLP.segment(raw) #保存清洗后的数据 wordslist1=str(wordslist).split(",") # print(wordslist1[1:len(wordslist1)-1]) flagresult="" # 去除标签 for v in wordslist1[1:len(wordslist1)-1]: if "/" in v: slope=v.index("/") letter=v[1:slope] if len(letter)>0 and '\n\u3000\u3000' in letter: flagresult+="\n" else:flagresult+=letter +"/" #去除空格 # print(flagresult) with open(os.path.join(save_folder_path,file),'w',encoding='utf-8') as f: f.write(flagresult.replace(' /','')) j += 1 shutdownJVM() if __name__ == '__main__' : print('开始进行文本分词操作:\n') t1 = time.time() dealpath="../Database/SogouC/FileTest/" savepath="../Database/SogouCCut/FileTest/" # 待分词的语料类别集根目录 read_folder_path = '../Database/SogouC/FileNews/' write_folder_path = '../Database/SogouCCut/' #jieba中文分词 CHSegment(read_folder_path,write_folder_path) #300个txtq其中结巴分词使用3.31秒 HanLPSeg(read_folder_path,write_folder_path) #300个txt其中hanlp分词使用1.83秒 t2 = time.time() print('完成中文文本切分: '+str(t2-t1)+"秒。") |
文章来源于网络
全部0条评论
快来发表一下你的评论吧 !