计算两个文档每一行的字错率cer

distance库的distance.levenshtein( , )函数可以计算两句话之间的距离（错+漏+缺）

下面代码是对比语音识别结果和标准结果之间的错误率。每一行进行对比，输出结果为

输入文件格式为：spk##啊啊啊

import distance
import re

def spk_compare(compare_spk_list,cor_spk_list):
    spk_compared_list=[]
    for spk_num in range(len(xf_spk_list)):
        if compare_spk_list[spk_num]==cor_spk_list[spk_num]:
            spk_compared_list.append('1')
        else:
            spk_compared_list.append('0')
    return spk_compared_list

def split_spk_content(list):
    spk_list=[]
    content_list=[]

    for line in list:
        if line=='\n':
            spk='spk0'
            content=''
        else:
            spk=(line.split('##'))[0]
            #print(len(line.split('#')))
            #print(line.split('#'))
            content=re.sub('[^\u4e00-\u9fa5]+', '',(line.split('##'))[1])#删除非中文字符
        spk_list.append(spk)
        content_list.append(content)
    return spk_list,content_list



def count_num(content_list):
    num_word_list=[]
    for line in content_list:
        num_word_list.append(len(line))

    # print(num_word_list)
    return num_word_list


def cal_distance(cort_sentences_list,compare_sentences_list):
    distance_list=[]
    for num_list in range(len(cort_sentences_list)):
        sentence_distance = distance.levenshtein(cort_sentences_list[num_list],compare_sentences_list[num_list])
        distance_list.append(sentence_distance)
    return distance_list

def x(content_list):
    x_list=[]
    flag='0'
    for line in content_list:
        for i in line:
            if i=="x":
                flag='1'
        x_list.append(flag)
        flag = '0'
    return x_list



if __name__ == '__main__':
    xunfei_txt = open('new.txt','r',encoding='utf-8')
    correct_txt = open('correct.txt','r',encoding='utf-8')
    xunfei_list=xunfei_txt.readlines()
    correct_list=correct_txt.readlines()
    x_list = x(correct_list)
    xf_spk_list,xf_content_list=split_spk_content(xunfei_list)
    cor_spk_list,cor_content_list=split_spk_content(correct_list)

    spk_compared_list=spk_compare(xf_spk_list,cor_spk_list)
    cor_num_word_list = count_num(cor_content_list)
    xf_num_word_list = count_num(xf_content_list)
    distance_list=cal_distance(xf_content_list,cor_content_list)
    # print(distance_list)

    print("x? spk?  字数correct 字数pred  dinstance  cer")
    for num in range(len(cor_spk_list)):

        print(x_list[num],spk_compared_list[num],"    ",cor_num_word_list[num],xf_num_word_list[num],distance_list[num],str(distance_list[num]/(cor_num_word_list[num])))

 第一列x表示句子中是否有听不清的内容

第二列表示是否有说话者判断错，

第三四列为中文字符个数

第五列为错字数

distance/总字数=cer  （可能大于100%）

注意：两个文档必须行行对应