CoLA任务的数据增强方法

主要有四种语法错误：

①少词-A类

②错字母-B类

③少字母-C类

④换词-D类

我将每个类分为1,2,3个错误数量，这样共计12个小类：

# FOR 1A
import random
err_data_num = int(len(data)/2)
res = []
for i in range(40166):
    s = data[i].split(" ")
    word_count = len(s)-1
    print("word conut:",word_count)
    print("raw:",' '.join(s))
    if(word_count >= 3):
        local_error_1 = random.randint(1,word_count) # 错误位置
        print("Error position:",local_error_1)
        s.pop(local_error_1)
        if(local_error_1==word_count):
            s.append(".")
        print("dst:",' '.join(s))
        res.append(' '.join(s))
        print("")

# FOR 2A
import random
res = []
for i in range( 160667,180750):
    s = data[i].split(" ")
    word_count = len(s)-1
    print("word conut:",word_count)
    print("raw:",' '.join(s))
    if(word_count >= 3):
        local_error_1 = random.randint(1,word_count) # 错误位置
        print("Error position 1:",local_error_1)
        s.pop(local_error_1)
        if(len(s)-1 >= 3):
            local_error_2 = random.randint(1,len(s)-1) # 错误位置
            print("Error position 2:",local_error_2)
            s.pop(local_error_2)
            if(local_error_2==len(s)-1):
                s.append(".")
        print("dst:",' '.join(s))
        res.append(' '.join(s))

# FOR 3A
import random
res = []
for i in range(241001,251042):
    s = data[i].split(" ")
    word_count = len(s)-1
    print("word conut:",word_count)
    print("raw:",' '.join(s))
    if(word_count >= 3):
        local_error_1 = random.randint(1,word_count) 
        print("Error position 1:",local_error_1)
        s.pop(local_error_1)
        if(len(s)-1 >= 3):
            local_error_2 = random.randint(1,len(s)-1) 
            print("Error position 2:",local_error_2)
            s.pop(local_error_2)
            if(len(s)-1 >= 3):
                local_error_3 = random.randint(1,len(s)-1) 
                print("Error position 3:",local_error_3)
                s.pop(local_error_3)
                if(local_error_3==len(s)-1):
                    s.append(".")
        print("dst:",' '.join(s))
        res.append(' '.join(s))

# FOR 1B
import random
res = []
vocab = ["a","b","c","d","e","f","g","h","i","j","k","l","m","n","o","p","q","r","s","t","u","v","w","x","y","z"]
for i in range(40167,80333):
    s = data[i].split(" ")
    word_count = len(s)-1
    if(word_count >= 1):
        danci_weizhi_1 = random.randint(1,word_count) # 错词位置
        ci = s[danci_weizhi_1]    # 取词
        print("raw:",ci)
        zimu_list = list(ci)      # 序列化
        local_zimu_error = random.randint(0,len(zimu_list)-1)  # 错字母位置
        random_vocab = random.randint(0,25)  
        zimu_list[local_zimu_error] = vocab[random_vocab]   # 替换
        tmp = ''.join(zimu_list)
        print("dst:",tmp)
        s[danci_weizhi_1] = tmp
        res.append(' '.join(s))

# FOR 2B
import random
res = []
vocab = ["a","b","c","d","e","f","g","h","i","j","k","l","m","n","o","p","q","r","s","t","u","v","w","x","y","z"]
for i in range(180751,200834):
    s = data[i].split(" ")
    word_count = len(s)-1
    if(word_count >= 1):
        #### 第一个
        danci_weizhi_1 = random.randint(1,word_count) # 错词位置
        ci = s[danci_weizhi_1]    # 取词
        print("raw:",ci)
        zimu_list = list(ci)      # 序列化
        local_zimu_error = random.randint(0,len(zimu_list)-1)  # 错字母位置
        random_vocab = random.randint(0,25)  
        zimu_list[local_zimu_error] = vocab[random_vocab]   # 替换
        tmp = ''.join(zimu_list)
        print("dst:",tmp)
        s[danci_weizhi_1] = tmp
        #### 第二个
        danci_weizhi_2 = random.randint(1,word_count) # 错词位置
        ci = s[danci_weizhi_2]    # 取词
        print("raw:",ci)
        zimu_list = list(ci)      # 序列化
        local_zimu_error = random.randint(0,len(zimu_list)-1)  # 错字母位置
        random_vocab = random.randint(0,25)  
        zimu_list[local_zimu_error] = vocab[random_vocab]   # 替换
        tmp = ''.join(zimu_list)
        print("dst:",tmp)
        s[danci_weizhi_2] = tmp
        
        res.append(' '.join(s))

# # FOR 3B
import random
res = []
vocab = ["a","b","c","d","e","f","g","h","i","j","k","l","m","n","o","p","q","r","s","t","u","v","w","x","y","z"]
for i in range(251043,261083):
    s = data[i].split(" ")
    word_count = len(s)-1
    if(word_count >= 1):
        #### 第一个
        danci_weizhi_1 = random.randint(1,word_count) # 错词位置
        ci = s[danci_weizhi_1]    # 取词
        print("raw:",ci)
        zimu_list = list(ci)      # 序列化
        local_zimu_error = random.randint(0,len(zimu_list)-1)  # 错字母位置
        random_vocab = random.randint(0,25)  
        zimu_list[local_zimu_error] = vocab[random_vocab]   # 替换
        tmp = ''.join(zimu_list)
        print("dst:",tmp)
        s[danci_weizhi_1] = tmp
        #### 第二个
        danci_weizhi_2 = random.randint(1,word_count) # 错词位置
        ci = s[danci_weizhi_2]    # 取词
        print("raw:",ci)
        zimu_list = list(ci)      # 序列化
        local_zimu_error = random.randint(0,len(zimu_list)-1)  # 错字母位置
        random_vocab = random.randint(0,25)  
        zimu_list[local_zimu_error] = vocab[random_vocab]   # 替换
        tmp = ''.join(zimu_list)
        print("dst:",tmp)
        s[danci_weizhi_2] = tmp
        #### 第三个
        danci_weizhi_3 = random.randint(1,word_count) # 错词位置
        ci = s[danci_weizhi_3]    # 取词
        print("raw:",ci)
        zimu_list = list(ci)      # 序列化
        local_zimu_error = random.randint(0,len(zimu_list)-1)  # 错字母位置
        random_vocab = random.randint(0,25)  
        zimu_list[local_zimu_error] = vocab[random_vocab]   # 替换
        tmp = ''.join(zimu_list)
        print("dst:",tmp)
        s[danci_weizhi_3] = tmp
        res.append(' '.join(s))

#FOR 1C
import random
res = []
for i in range(80334,120500):
    s = data[i].split(" ")
    word_count = len(s)-1
    if(word_count >= 1):
        danci_weizhi_1 = random.randint(1,word_count) # 错词位置
        ci = s[danci_weizhi_1]    # 取词
        print("raw:",ci)
        zimu_list = list(ci)      # 序列化
        local_zimu_error = random.randint(0,len(zimu_list)-1)  # 错字母位置
        zimu_list.pop(local_zimu_error)
        tmp = ''.join(zimu_list)
        print("dst:",tmp)
        s[danci_weizhi_1] = tmp
        res.append(' '.join(s))

# #FOR 2C
import random
res = []
for i in range(200835,220918):
    s = data[i].split(" ")
    word_count = len(s)-1
    if(word_count >= 2):
        danci_weizhi_1 = random.randint(1,word_count) # 错词位置
        ci = s[danci_weizhi_1]    # 取词
        print("raw:",ci)
        zimu_list = list(ci)      # 序列化
        local_zimu_error = random.randint(0,len(zimu_list)-1)  # 错字母位置
        zimu_list.pop(local_zimu_error)
        tmp = ''.join(zimu_list)
        print("dst:",tmp)
        s[danci_weizhi_1] = tmp
        
        danci_weizhi_2 = random.randint(1,word_count) # 错词位置
        ci2 = s[danci_weizhi_2]    # 取词
        print("raw:",ci2)
        zimu_list2 = list(ci2)      # 序列化
        if(len(zimu_list2)-1 >0):
            local_zimu_error2 = random.randint(0,len(zimu_list2)-1)  # 错字母位置
            zimu_list2.pop(local_zimu_error2)
            tmp2 = ''.join(zimu_list2)
            print("dst:",tmp2)
            s[danci_weizhi_2] = tmp2
        res.append(' '.join(s))

# #FOR 3C
data = []
for line in open("data4.txt","r"): 
    data.append(line)
len(data)
import random
res = []
for i in range(261084,271124):
    s = data[i].split(" ")
    word_count = len(s)-1
    if(word_count >= 2):
        danci_weizhi_1 = random.randint(1,word_count) # 错词位置
        ci = s[danci_weizhi_1]    # 取词
        print("raw:",ci)
        zimu_list = list(ci)      # 序列化
        if(len(zimu_list)-1 > 0):
            local_zimu_error = random.randint(0,len(zimu_list)-1)  # 错字母位置
            zimu_list.pop(local_zimu_error)
            tmp = ''.join(zimu_list)
            print