如何用Python库拆分DataFrame文本列生成包含/排除列？

思路 #

观察文本的固定模式：所有条目均以includes系列短语引出包含内容，随后通过but/however等转折词衔接doesn’t include引出排除内容。这种结构化文本无需复杂NLP模型，用正则表达式即可精准提取目标内容。

实现代码 #

1. 准备示例数据

import pandas as pd

data = {
    "column_description": [
        "this section includes: animals: cats and dogs and vegetables but doesn’t include: plants and fruits: coco",
        "this section includes the following: axis: x and y but doesn’t include: z, k and c",
        "this section includes notably: letters: a, b and c however it doesn’t include: y and letter: z"
    ]
}
df = pd.DataFrame(data)

2. 提取包含/排除内容

# 提取column_include：匹配includes后到转折词前的内容
df["column_include"] = df["column_description"].str.extract(
    r"includes.*?: (.*?) (but|however)", expand=False
)[0].str.strip()

# 提取column_exclude：匹配doesn’t include后的所有内容
df["column_exclude"] = df["column_description"].str.extract(
    r"doesn’t include: (.*)$", expand=False
).str.strip()

3. 查看结果

print(df[["column_include", "column_exclude"]])

输出结果：

column_include          column_exclude
0  animals: cats and dogs and vegetables  plants and fruits: coco
1                      axis: x and y           z, k and c
2           letters: a, b and c        y and letter: z

正则说明 #

• includes.*?: (.*?) (but|however)：
  • includes.*?: 非贪婪匹配从includes到最后一个冒号的内容（适配不同表述如includes the following:）
  • (.*?) 捕获冒号后到转折词前的目标文本
• doesn’t include: (.*)$：捕获doesn’t include:后到文本结尾的所有内容

扩展适配 #

如果文本中出现其他转折词（如yet），只需修改正则中的转折词列表：(but|however|yet)；若存在无排除内容的条目，可添加fillna("")处理空值：

df["column_exclude"] = df["column_exclude"].fillna("")

内容的提问来源于stack exchange，提问作者Paradisum