有一个如下数据框。
id num text
1 1.2 price is 1.2
1 2.3 price is 1.2 or 2.3
2 3 The total value is $3 and $130
3 5 The apple value is 5dollar and $150
我想用字符“ UNK”替换文本中的num
,新的数据框更改为:
id num text
1 1.2 price is UNK
1 2.3 price is 1.2 or UNK
2 3 The total value is UNK and 130
3 5 The apple value is UNK dollar and $150
z 我当前的代码如下
df_dev['text'].str.replace(df_dev['num'], 'UNK')
有错误:
TypeError: 'Series' objects are mutable, thus they cannot be hashed
答案 0 :(得分:2)
让我们使用regex
和replace
df.text.replace(regex=r'(?i)'+ df.num.astype(str),value="UNK")
0 price is UNK
1 price is 1.2 or UNK
2 The total value is UNK
Name: text, dtype: object
#df.text=df.text.replace(regex=r'(?i)'+ df.num.astype(str),value="UNK")
更新
(df.text+' ').replace(regex=r'(?i) '+ df.num.astype(str)+' ',value=" UNK ")
0 price is UNK
1 price is 1.2 or UNK
2 The total value is UNK and 130
Name: text, dtype: object
答案 1 :(得分:1)
该错误是正确的,您无法向expects a string or regular expression的方法提供序列。
Pandas字符串方法不是矢量化的,即它们在后台涉及Python级循环,因此列表理解可能会很好地工作:
zipper = zip(df['text'], df['num'].astype(str))
df['text'] = [text.replace(num, 'UNK') for text, num in zipper]