将部分文本与全文匹配并替换

Posted 2023-02-22

【中文标题】将部分文本与全文匹配并替换

【英文标题】：Match partial text with full text and replace

【发布时间】：2021-10-04 02:49:12

【问题描述】：

我有一个推文数据集，其中一些推文是原创的，而其他推文是转发的。由于某种原因，转推被... 截断，因此整个文本不存在。在我的数据集中，原始推文（希望）始终存在，所以我想找到原始推文并用它替换截断的推文。

例如：

my_data <- tribble(
  ~user, ~text,
  "Peter", "Hello, this is Peter, I like ice cream!",
  "John", "RT @Peter: Hello, this is Peter, I like ...",
  "Martha", "RT @Peter: Hello, this is Peter, I like ...",
  "Julia", "Hi, I really like apples!",
  "Bjorn", "RT @Julia: I really like ..."
)

# A tibble: 5 x 2
  user   text                                       
  <chr>  <chr>                                      
1 Peter  Hello, this is Peter, I like ice cream!    
2 John   RT @Peter: Hello, this is Peter, I like ...
3 Martha RT @Peter: Hello, this is Peter, I like ...
4 Julia  Hi, I really like apples!                  
5 Bjorn  RT @Julia: I really like ... 

我想找到RT@ username: some text... 的每个实例，并将其替换为完整的推文。基本上：

# A tibble: 5 x 2
  user   text                                              
  <chr>  <chr>                                             
1 Peter  Hello, this is Peter, I like ice cream!           
2 John   RT @Peter: Hello, this is Peter, I like ice cream!
3 Martha RT @Peter: Hello, this is Peter, I like ice cream!
4 Julia  Hi, I really like apples!                         
5 Bjorn  RT @Julia: Hi, I really like apples!     

我已经提取了被转发的句柄，并将其按组分解：

retweet_pattern <- "^RT @([a-zA-Z0-9_]*): (.*)"
str_match(my_data$text, retweet_pattern)

但是，我不完全确定如何进行。由于用户/文本对不一定是唯一的（即，一个用户可能有多个被转发的推文），简单地找到一个转发句柄并更改整个文本是行不通的。也许我需要使用字符串指标，比如 Levenshtein？

谢谢。

【参考方案1】：

由于转推文本与非转推数据完全一致，你可以试试这个。

library(dplyr)
library(tidyr)

#Create a separate dataframe for retweet data
#separate the username and tweet in different columns
rt_data <- my_data %>% 
  filter(grepl('RT', text)) %>%
  separate(text, c('name', 'text'), sep = ':\\s*')

#Create a separate dataframe for tweets which are not retweets. 
no_rt_data <- my_data %>% filter(!grepl('RT', text))
  

#Clean the retweet string and find the corresponding match
#in non-retweet data
rt_data$text <- sapply(gsub('RT @\\w+:\\s*|\\s*\\.+$', '', rt_data$text), 
                       function(x) no_rt_data$text[grepl(x, no_rt_data$text)])

#Combine the username and tweet  
rt_data <- rt_data %>% unite(text, name, text, sep = ' :')

#combine the two dataframes
bind_rows(no_rt_data, rt_data)

#   user   text                                              
#  <chr>  <chr>                                             
#1 Peter  Hello, this is Peter, I like ice cream!           
#2 Julia  Hi, I really like apples!                         
#3 John   RT @Peter :Hello, this is Peter, I like ice cream!
#4 Martha RT @Peter :Hello, this is Peter, I like ice cream!
#5 Bjorn  RT @Julia :Hi, I really like apples!