如何在R语言数据框中提取指定键值对至独立列

可以借助dplyr、stringr和tidyr包处理这类非结构化键值文本，核心思路是拆分文本、识别键值、转换为结构化格式：

代码实现 #

首先加载用户提供的数据：

# 加载数据
asummy <- structure(list(`arrange_df[, 12]` = c("Customer Name: JOE\nReason for the Call: BP Set Up\nResolution: \nvisual audit on the account\nset expectations in BP\noffered call back\nCenter Location: Clark\nCTN (Number Calling About): ************************************\nAVAYA (Number Calling From): ************************************\nAgent UID: cm***************************u\n", 
"Name: Kim Putok\nCtn: ************************************\nReason for calling: lost phone/ ************************************/ follow up on insurance claim/ routed to asurion\nResolution:\nLocation: Clark\nattuid: gb***************************r\n", 
"Customer Name: Heather \nReason for the Call:    got suspended / card issue / supposedly paid / complains she just updated online her card then got susp\nResolution:  \nCenter Location: Clark\nCTN:  ************************************\nAlt No:  ************************************\nCredit Reason:  ********* courtesy \nAgent UID: rc***************************w/I-EQR******************G\n", 
"Customer Name: GLORIA ;      CTN: ************************************      Affected CTN: ************************************      Alternate Number: none      Reason for the Call: change rate plan      Resolution: set exp on changing rate plan      Agent UID: gt***************************g", 
"Customer Name: BRANDY \nReason for the Call: payment\nRecommendations/Troubleshooting Steps:\nResolution: explained card errors\nasked for alt # but no good\noffered to use different card or refill card\nsent qp link\nshe will go to bank to check as per cx\nCenter Location:\nCTN: ************************************\nCredit Reason (If credit was applied to account):\nAgent UID: jq***************************n\n", 
"Customer Name: ; ;Leahi ;  Reason for the Call: ; ; ;billing issue / ******************.****************** dollars only orig. bill / due date shld be *********th / why ****************** now? / ctn change questions ;  Resolution: ; ;explained bill / ctn change info provided ;  Center Location: ;Clark  CTN: ; ;************************************  Alt No:  Credit Reason:  Agent UID: rc***************************w/I-K*********GMDR", 
"Customer Name: MICHAEL\nReason for the Call: ACCOUNT BAL INQ\n\nRecommendations/Troubleshooting Steps:\n*Account verified and provided information\n\nResolution: \n*calling about the account status\n*suspended due to BBP\n*adv about BP policy\n*Provided payment options\n\nCenter Location:Clark\nCTN************************************:\nCredit Reason (If credit was applied to account):\nAgent UID:rc*********\n", 
"Customer Name: Kimberly\nCTN: ************************************\nAffected CTN:************************************\nAlternate Number: none\nReason for the Call: add mhs otc $******************\nResolution: added mhs otc  $******************\nAgent UID: gt***************************g\n", 
"auto pop / suspended line\n-cx speaking spanish\n-transfered call\n", 
"Customer Name: TERRELL ;  Reason for the Call: payment  Recommendations/Troubleshooting Steps:  Resolution: payment success  test and validated  CTN: ************************************  Credit Reason (If credit was applied to account):  Agent UID: jq***************************n"
)), row.names = c(NA, 10L), class = "data.frame")

然后提取所有键值对：

library(dplyr)
library(stringr)
library(tidyr)

asummy %>%
  mutate(row_id = row_number(),  # 保留行号，避免丢失文本对应关系
         text = str_replace_all(`arrange_df[, 12]`, "\\s*;\\s*", " "),  # 清理冗余分号和空格
         split_text = str_split(text, "(?<=:)\\s*|\\n")) %>%  # 按换行或冒号后边界拆分文本
  unnest(split_text) %>%
  filter(split_text != "") %>%  # 过滤空条目
  mutate(
    key = str_extract(split_text, "^[^:]+(?=:)"),  # 提取键名
    value = str_trim(str_replace(split_text, "^[^:]+:\\s*", "")),  # 提取对应值
    key = str_trim(key)
  ) %>%
  filter(!is.na(key)) %>%  # 过滤无法识别为键值对的内容
  pivot_wider(names_from = key, values_from = value, values_fn = ~paste(., collapse = "\n")) %>%
  select(-text)

该方法会自动识别所有键名: 值格式的内容，处理多行值、空格冗余等问题，最终将每个键转为独立列。

针对特定字段，用精准正则匹配兼容字段名变体（如Reason for calling），同时保留多行内容：

代码实现 #

asummy %>%
  mutate(
    # 提取Reason字段，兼容"Reason for the Call"和"Reason for calling"写法
    `Reason for the Call` = str_extract(`arrange_df[, 12]`, "(?<=Reason for (the )?call:).*?(?=\\n[^\\s]|$)"),
    # 提取Resolution字段，捕获到下一个键名或文本结尾的所有内容
    Resolution = str_extract(`arrange_df[, 12]`, "(?<=Resolution:).*?(?=\\n[^\\s]|$)"),
    # 清理多余换行、分号和首尾空格
    across(c(`Reason for the Call`, Resolution), ~str_trim(str_replace_all(., "\\s*;\\s*|\\n+", " ")))
  ) %>%
  select(`Reason for the Call`, Resolution)

说明 #

• 正则(?<=Reason for (the )?call:)用正向断言匹配不同写法的Reason字段名
• .*?(?=\\n[^\\s]|$)确保捕获完整的字段值（包括多行内容），直到下一个非空白开头的行（即下一个键名）
• 对无目标字段的行，结果显示为NA，符合预期

内容的提问来源于stack exchange，提问作者Rich