如何根据datafram中的条件将行与前一行合并

Question

我有一个数据框，其中每一行都是一个单词或标点符号。我想遍历数据帧，每当一行包含标点符号时，我就想把它和前一行结合起来。你知道吗

例如，我想转换：

word 0 hello 1 , 2 how 3 are 4 you 5 ?

Answer 1

另一种方法，使用.shift(-1)连接到前一行：

df.loc[df["word"].shift(-1).isin(list(punctuation)), "word"] = df["word"] + df["word"].shift(-1)
df = df[~df["word"].isin(list(punctuation))][["word"]]

测向：

     word
0  hello,
2     how
3     are
4    you?

Answer 2

您可以使用isin和cumsum：

# list of puctuations
punctuations = set([',','?']) 

# blocks
blocks = ~df['word'].isin(punctuations)).cumsum()

# groupby
df['word'].groupby(blocks).sum()

输出：

word
1    hello,
2       how
3       are
4      you?
Name: word, dtype: object

Answer 3

match和cumsum

df.groupby((~df.word.str.match('\W')).cumsum(), as_index=False).sum()

     word
0  hello,
1     how
2     are
3    you?

---

isin

而且，没有as_index=True

from string import punctuation

df.groupby((~df.word.isin(list(punctuation))).cumsum()).sum()

        word
word        
1     hello,
2        how
3        are
4       you?