如何用R语言按规则从多段文本提取含指定词汇的句子

需求 #

• 从多行文本中提取符合以下规则的句子：
  1. 包含单词bonus或incentive（不区分大小写）；
  2. 句子以标点（. ! ?）、换行或控制字符分隔。

测试数据 #

text <-  c("This is a sentence. $5k SIGN-ON BONUS offered. This is another sentence. Salary is $15.00 per hours. Another",
         "This is a sentence. Retention bonus of $5,000 offered! This is another sentence. Salary is $15.00 per hours? Another", 
         "This is a sentence. $5k incentive offered! This is another sentence. Salary is $15.00 per hours. Another", 
         "This is a sentence\n \n$5000 sign-on Bonus offered\n \nThis is another sentence\n \nSalary is $15.00 per hours\n \nAnother", 
         "This is a sentence\n\nRetention bonus of $5000 offered\n\nThis is another sentence\n\nSalary is $15.00 per hours\n\nAnother",
         "This is a sentence\n \n$5k incentive offered\n \nThis is another sentence\n Salary is $15.00 per hours\nAnother",
         
         "This is a sentence. 
          $5k signing bonus offered! 
          This is another sentence. 
          Salary is $15.00 per hours? Another", 
         
         "This is a sentence. 
          
          This is another sentence. 
          
          $5k incentive offered! 
          Salary is $15.00 per hours? Another")

之前的尝试及问题 #

使用stringr::str_extract的正则表达式未得到理想结果：

stringr::str_extract(text, "[[:print:]]*(?i)bonus|(?i)incentive[[:print:]]*[[:cntrl:]]|[[:punct:]]")

[1] "This is a sentence. $5k SIGN-ON BONUS" "This is a sentence. Retention bonus"  
[3] "."                                     "$5000 sign-on Bonus"                  
[5] "Retention bonus"                       "incentive offered\n"                  
[7] "."                                     "."

期望输出：

[1] "$5k SIGN-ON BONUS offered"                "Retention bonus of $5,000 offered"  
[3] "$5k incentive offered"                    "$5000 sign-on Bonus offered"                  
[5] "Retention bonus of $5000 offered"         "$5k incentive offered"                  
[7] "$5k signing bonus offered"                "$5k incentive offered"

解决方案 #

之前的正则逻辑未精准界定句子边界，导致提取结果混乱。以下两种方法可解决问题：

方法一：拆分句子后筛选（逻辑清晰） #

先将文本按句子分隔符拆分为独立句子，再筛选包含目标关键词的句子：

library(stringr)

# 定义句子分隔规则：标点后接空白/控制字符，或直接控制字符
split_pattern <- "(?<=[.!?])\\s*|\\s*[[:cntrl:]]\\s*"

# 遍历每个文本元素，拆分、筛选、清理空白
result <- sapply(text, function(x) {
  sentences <- str_split(x, split_pattern)[[1]]
  target_sentences <- str_subset(sentences, regex("bonus|incentive", ignore_case = TRUE))
  str_trim(target_sentences)
}, USE.NAMES = FALSE)

# 整理为一维向量
result <- unlist(result)
print(result)

方法二：正则直接提取（一步到位） #

调整正则表达式，用断言精准匹配句子的前后边界：

library(stringr)

# 匹配包含关键词的完整句子，忽略大小写
extract_pattern <- regex("(?<=^|\\s|[.!?]|\\n)\\s*([^.!?\\n]+(bonus|incentive)[^.!?\\n]+)\\s*(?=[.!?]|\\n|$)", ignore_case = TRUE)

result <- str_extract(text, extract_pattern)
# 清理句子前后的空白字符
result <- str_trim(result)
print(result)

输出验证 #

两种方法均可得到符合预期的输出：

[1] "$5k SIGN-ON BONUS offered"                "Retention bonus of $5,000 offered"  
[3] "$5k incentive offered"                    "$5000 sign-on Bonus offered"                  
[5] "Retention bonus of $5000 offered"         "$5k incentive offered"                  
[7] "$5k signing bonus offered"                "$5k incentive offered"

逻辑说明 #

• 拆分法：先拆分再筛选，逻辑直观，避免正则边界匹配的复杂问题；
• 直接提取法：通过正向/反向断言锁定句子的起始和结束位置，确保只提取包含关键词的完整句子。

内容的提问来源于stack exchange，提问作者nebulous