如何在保持列不变的同时使用列表元素填充 Pandas 数据框？

如何解决如何在保持列不变的同时使用列表元素填充 Pandas 数据框？

我有一个 Pandas 数据框 df，其中包含许多行和 2 列，如下所示：

| Query | Description |
| -------- | -------------- |
| First sentence   |  First description           |
| Second sentence  | Second description            |

我创建了一个方法 new_sentences(query)，它接受​​一个 Query 句子并从中生成 n 个更相关的句子。目标是创建一个由原始数据框+通过new_sentences(query) 方法获得的新句子组成的新数据框。这个方法返回一个新句子的列表，给定一个查询，例如对于原始数据框中的第一句话 ["Generation 1 for first sentence","Generation 2 for first sentence",..."Generation n for first sentence"] 等 其他传递的句子也有类似的成立（尽管返回的列表可以为每个传递的 Query 包含不同数量的元素）。

对于相同类型的查询对应的新句子，描述应该保持不变。例如：

for i in range(len(df)):
  new_sentences = new_sentences(str(df['Query'].iloc[i]))
  #Append the new_sentences but keep description same.

预期的输出是这样的：

| Query | Description |
| -------- | -------------- |
| First sentence   |  First description           |
| Generation 1 for first sentence   |  First description           |
| Generation 2 for first sentence   |  First description           |
| Second sentence  | Second description            |
| Generation 1 for second sentence   |  Second description           |
| Generation 2 for second sentence   |  Second description           |
....

等

所以这个想法是用这些新的查询语句填充数据框（除了已经存在的旧语句），以便有一个更大的数据框。我尝试采用循环方法并使用 iterrows 对数据框进行迭代，但我不确定如何使新生成的句子的描述保持相同。任何帮助和指导/建议表示赞赏。谢谢！

解决方法

获取示例数据帧

df = pd.DataFrame([['hello','how'],['are','you']],columns=['Query','Description'])
df
    Query   Description
0   hello   how
1   are     you  

并采用一个虚拟的 new_sentences 函数，您可以在其中将新句子与原始句子一起作为列表返回。

def new_sentences(x):
    return [x['Query'],x['Query'][::-1]]

然后你可以沿着行（轴=1）使用apply函数

df['Query'] = df.apply(new_sentences,axis=1)
df
        Query       Description
0   [hello,olleh]  how
1   [are,era]      you

现在，您只需 explode 列表并进行所需的格式设置

(df.set_index('Description')['Query'] # setting temporary index for exploding
 .explode().reset_index()
 .reindex(['Query','Description'],axis=1)) # ordering the columns

    Query   Description
0   hello   how
1   olleh   how
2   are     you
3   era     you

,

您也可以通过转换为列表来实现：

df1 = pd.DataFrame({'Query': ['how healthy are you','how wealthy is your father'],'Description': ['question about you','question about your dad']})

#a sample function that slices each query to produce two new queries
def new_sentences(query):
    return [query[:11],query[12:]]

old_queries = df1['Query'].tolist()
descrips = df1['Description'].tolist()
df2 = pd.DataFrame([[new,descrips[i]] for i,q in enumerate(old_queries)
                    for new in new_sentences(q)],'Description'])

print(df1)
print(df2)

#original df1:
                        Query              Description
0         how healthy are you       question about you
1  how wealthy is your father  question about your dad 

#new df2
            Query              Description
0     how healthy       question about you
1         are you       question about you
2     how wealthy  question about your dad
3  is your father  question about your dad