linux 上安装CRF++ 、问题和代码

1.在Linus 上安装crf++

1）比较常用的工具就是CRF++。官网地址为：https://taku910.github.io/crfpp/ 选择CRF+±0.58.tar.gz

2）tar zxvf CRF+±0.58.tar.gz 进入CRF+±0.58目录下

3) ./configure

4)make

5)su

6)make install

注意：这里用crf++ 工具跑出的数据都是在所在安装包目录下跑的数据。不知在其他目录可以访问：

/home/yu/crf/CRF+±0.58/crf_test -m model test.data >> tr.txt

以上测试后没有问题。

安装是否成功，测试一下：

1）用CRF++ -0.58 包里面的模板跑出模型 model

2. 用跑出来的模型 model 去跑测试数据 test.data ，>> test.txt 意思是跑出的测试数据写入到 test.txt

3. 查看跑出的数据test.txt 文件

2.关于训练中文数据遇到的问题

1.遇到用cat 打开文件显示乱码，首先把文件转换成utf-8 的无 BOM 的格式代码。

2.用model 跑测试数据出现，只训练最后一段话，前面都没啦，报这样错误，不太懂，但是换了测试文件后，就可以啦。

3.当要写入一个以当前时间为名的txt 文档，可以字符串之间组合，不用犹豫不决。代码如下：

result_file = datetime.now().strftime("%Y%m%d%H%M%S") + '_' +'.txt'
print  result_file

输出：

• 20170528211827_.txt

3.代码：

#-*- coding:utf-8 -*-


import sys

import subprocess
from datetime import datetime,timedelta
import random ,string


reload(sys)
sys.setdefaultencoding('utf-8')



def runLinuxCommandOnPY(path_crf_test,path_model,path_test,path_output,path_result):
    flag_success = 1
    p = subprocess.Popen(path_crf_test+' -m '+path_model+' '+ path_test,
                         shell=True, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
    a = list(string.ascii_letters)
    random.shuffle(a)
   
    path_write_test = path_output+''.join(a[:4])+'.txt'
    f_trained = open(path_write_test,'w')
    for line in p.stdout.readlines():
        f_trained.write(line.strip()+'\n')
        print line,
    f_trained.close()
    etval = p.wait()
    result_file = datetime.now().strftime("%Y%m%d%H%M%S")+'_'+''.join(a[:2])+'.txt'
    
	
    if etval == 0:
        flag_success = 0

    if flag_success ==0:
        test_zhengwen = open(path_write_test, 'r')
        s_zhengwen = test_zhengwen.readlines()
        f = open(path_result+result_file, 'w')

        str_t = ''
        str_ns = ''
        str_nr = ''
        str_nt = ''
        
        for i in s_zhengwen:

            if (i.strip()):
                m = i.strip().split('\t')
               
                if m[2] == 'S':
                    f.write(m[0])
                else:
                    if m[1] == 't':
                        
                        if m[2] == 'B' or m[2] == 'M':
                            str_t += m[0]
                        else:
                            str_t += m[0]
                            st = '{{time:' + str_t + '}}'
                            f.write(st)
                            str_t = ''
                    elif m[1] == 'ns':
    
                        if m[2] == 'B' or m[2] == 'M':
                            str_ns += m[0]
                        else:
                            str_ns += m[0]
                            sns = '{{location:' + str_ns + '}}'
                            f.write(sns)
                            str_ns = ''
                    elif m[1] == 'nr':
                        
                        if m[2] == 'B' or m[2] == 'M':
                            str_nr += m[0]
                        else:
                            str_nr += m[0]
                            snr = '{{person_name:' + str_nr + '}}'
                            f.write(snr)
                            str_nr = ''
                    elif m[1] == 'nt':
                        
                        if m[2] == 'B' or m[2] == 'M':
                            str_nt += m[0]
                        else:
                            str_nt += m[0]
                            snt = '{{org_name:' + str_nt + '}}'
                            f.write(snt)
                            str_nt = ''
                    else:
                       
                        f.write(m[0])
        f.close()
    subprocess.call("rm -rf "+path_write_test, shell=True)
    
    return path_result+result_file

调用：

#-*- coding:utf-8 -*-


import sys
import executeLinuxCommand

reload(sys)
sys.setdefaultencoding('utf-8')

def test():
    path_crf_test = '../CRF++-0.58/crf_test'
    path_model ='../entityRecognition/ceShi/yu/model'
    path_test = '../entityRecognition/ceShi/yu/gtest.data'
    path_output ='../entityRecognition/ceShi/yu/over_file/'
    path_result ='../entityRecognition/ceShi/yu/result/'
    result_file_path  = executeLinuxCommand.runLinuxCommandOnPY(path_crf_test,path_model,
                                            path_test,path_output,path_result)
    print result_file_path 
if __name__ == '__main__':
    test()

安装过程：