时间戳

首页 > 解决方案 > 将对话小标题转换为 .txt，然后再返回

r - 将对话小标题转换为 .txt，然后再返回

问题描述

我想取一个代表对话的 tibble 并将其转换为可以在文本编辑器中手动编辑的 .txt，然后返回到 tibble 进行处理。

我遇到的主要挑战是以某种方式分隔文本块，以便在编辑后可以将它们重新导入为类似的格式，同时保留“发言人”的名称。

速度很重要，因为文件量和每个文本段的长度都很大。

这是输入小标题：

tibble::tribble(
    ~word, ~speakerTag,
   "been",          1L,
  "going",          1L,
     "on",          1L,
    "and",          1L,
   "what",          1L,
   "your",          1L,
  "goals",          1L,
   "are.",          1L,
  "Yeah,",          2L,
     "so",          2L,
     "so",          2L,
   "John",          2L,
    "has",          2L,
     "15",          2L
  )

这是 .txt 中所需的输出：

###Speaker 1###
been going on and what your goals are.
###Speaker 2###
Yeah, so so John has 15

这是手动更正错误后所需的回报：

    ~word, ~speakerTag,
   "been",          1L,
  "going",          1L,
     "on",          1L,
    "and",          1L,
   "what",          1L,
   "your",          1L,
  "goals",          1L,
   "in",            1L,
   "r",             1L,
  "Yeah,",          2L,
     "so",          2L,
     "so",          2L,
   "John",          2L,
    "hates",        2L,
     "50",          2L
  )

标签： rdplyrpurrrstringrgoogle-language-api

解决方案

一种方法是"\n"在每个开头添加演讲者姓名speakerTag

library(data.table)
library(dplyr)
library(tidyr)

setDT(df)[, word := replace(word, 1, paste0("\n\nSpeaker", 
            first(speakerTag), '\n\n', first(word))), rleid(speakerTag)]

我们可以在文本文件中使用

writeLines(paste(df$word, collapse = " "), 'Downloads/temp.txt')

它看起来像这样：

cat(paste(df$word, collapse = " "))

#Speaker1
#
#been going on and what your goals are. 
#
#Speaker2
#
#Yeah, so so John has 15

要在 R 中读回它，我们可以这样做：

read.table('Downloads/temp.txt', sep="\t", col.names = 'word') %>%
    mutate(SpeakerTag = replace(word, c(FALSE, TRUE), NA)) %>%
    fill(SpeakerTag) %>%
    slice(seq(2, n(), 2)) %>%
    separate_rows(word, sep = "\\s") %>%
    filter(word != '')

#    word SpeakerTag
#1   been   Speaker1
#2  going   Speaker1
#3     on   Speaker1
#4    and   Speaker1
#5   what   Speaker1
#6   your   Speaker1
#7  goals   Speaker1
#8   are.   Speaker1
#9  Yeah,   Speaker2
#10    so   Speaker2
#11    so   Speaker2
#12  John   Speaker2
#13   has   Speaker2
#14    15   Speaker2

Obviously we can remove "Speaker" part in SpeakerTag column if it is not needed.

推荐阅读