以下代码主要处理商品规格含有“每”的特殊处理（如：每袋5g，标准化为：5g/袋），默认处理方式可以参考上一篇随笔：https://www.jianshu.com/p/54a99c516c77

加载单位数据，比如：支，盒等，这里根据自己业务数据自行整理：

unit_data=pd.read_excel("unit.xlsx")

print("单位数据：",len(unit_data))

unit_set = set(unit_data.name)

定义关键字，用于特殊处理：

key_set = set(("装","重"))

根据业务判断进行连续字判断：

def is_chinese(c):
    return '\u4e00' <= c <= '\u9fff';

def char_test(c):
    return c.isdigit() or (c >= u'\u0041' and c <= u'\u005a') or (c >= u'\u0061' and c <= u'\u007a') or c == '.' or c == '/' or (c in unit_set)

标准化实现：

def specification_extract(txt):
    strs = []
    key_indexes = []
    begin = 0
    end = len(txt)
    index = txt.find("每",begin,end)
    
    if index < 0:
        strs.append(txt)
    else:
        if index > 0:
            key_indexes.append(0)
     
        while index > -1:
            key_indexes.append(index)
            begin = index + 1
            index = txt.find("每",begin,end)
        
    
        
        i = 0
    
        while i < len(key_indexes):
            if i + 1 < len(key_indexes):
                strs.append(txt[key_indexes[i]:key_indexes[i+1]])
            else:
                strs.append(txt[key_indexes[i]:])
        
            i = i+1
    
    return strs

def specification_normalize(txt):
    start_index = txt.find("每")
    
    #print(start_index)
    
    if start_index < 0:
        return txt
    
    unit_txt ="" 
    
    if is_chinese(txt[start_index+1]) == True:
        unit_txt = txt[start_index+1]
    
        #print(unit_txt)
    
        if txt[start_index+2] in key_set:
            start_index += 3
        else:
            start_index += 2
    else:
        start_index += 1
    
    i = start_index
    find = False
    find_index = 0
    
    #print(i,txt[start_index:])
    
    while i < len(txt) and find == False:
        find = (char_test(txt[i]) == False)
        
        #print(txt[i],find)
        
        if find == False:
            i += 1
        
    #print(start_index,i)
        
    result = txt[start_index:i] + "/" + unit_txt
    
    if i < len(txt)-1:
        result = result + txt[i:]
    
    return result

def specification_all_normalize(txt):
    if txt.find("每") < 0:
        return txt
    else:
        return "".join(map(specification_normalize,specification_extract(txt)))