Pandas合并连续行实现每行单词数≥10且不拆分原句子的实现方案

核心逻辑是为连续行生成专属分组标签：遍历每行累计单词数，一旦累计值≥设定的最低阈值（此处为10）就闭合当前分组，后续行进入新分组重新累计，最终按分组聚合即可实现需求，全程不会拆分原有完整句子。

完整实现代码 #

import pandas as pd

# 构造示例DataFrame
df = pd.DataFrame(data=[['Tyger Tyger burning.', 3],
                    ['bright.', 1],
                   ['In the forests of the night.', 6],
                   ['What immortal.', 2],  
                   ['hand or eye could frame thy fearful symmetry.',  8],
                   ['In what distant deeps or skies.', 6],
                   ['Burnt the fire of thine eyes.', 6]
                    ],columns=['SENTENCE','NO_WORDS'])

# 生成分组标签
current_word_sum = 0
group_id = 0
group_list = []
min_word_threshold = 10  # 可按需调整最低单词数阈值
for word_cnt in df['NO_WORDS']:
    current_word_sum += word_cnt
    group_list.append(group_id)
    # 累计单词数达标，开启新分组
    if current_word_sum >= min_word_threshold:
        group_id += 1
        current_word_sum = 0

# 按分组聚合得到结果
result_df = df.groupby(group_list).agg(
    SENTENCE = ('SENTENCE', ' '.join),
    NO_WORDS = ('NO_WORDS', 'sum')
).reset_index(drop=True)

print(result_df)

输出结果 #

SENTENCE  NO_WORDS
0  Tyger Tyger burning. bright. In the forests of...        10
1  What immortal. hand or eye could frame thy fea...        10
2  In what distant deeps or skies. Burnt the fire...        12

内容的提问来源于stack exchange，提问作者rumpet