formulate capture groups for inconsistently present substrings-CodePudding

I have transcriptions of interviews that are partly irregularly formed:

tst <- c("In: ja COOL;  #00:04:24-6#  ",           
         "  in den vier, FÜNF wochen, #00:04:57-8# ",
         "In: jah,  #00:02:07-8# ",
         "In:     [ja; ] #00:03:25-5# [ja; ] #00:03:26-1#",
         "    also jA:h; #00:03:16-6# (1.1)",
         "Bz:        [E::hm;    ]  #00:03:51-4#  (3.0)  ",
         "Bz:    [mhmh,      ]",
         "  in den bilLIE da war;")

What I need to do is structure this data by extracting its key elements into columns of a dataframe. There are four such key elements:

Rolein interview: interviewee or interviewer
Utterance: the interview partners' speech
Timestampindicated by # to both ends
Gap indicated by decimal number in brackets

The problem is that both Timestamp and Gapare inconsistently provided. While I can make the last capture group for Gap optional, those strings that have neither Timestamp nor Gapare not rendered correctly:

I'm using extract from tidyr for the extraction:

library(tidyr)
data.frame(tst) %>%
  extract(col = tst,
          into = c("Role", "Utterance", "Timestamp", "Gap"),
          regex = "^(\\w{2}:\\s|\\s )([\\S\\s] ?)\\s*#([^#] )?#\\s*(\\([0-9.] \\))?\\s*")
  Role                 Utterance  Timestamp   Gap
1 In:                   ja COOL; 00:04:24-6      
2      in den vier, FÜNF wochen, 00:04:57-8      
3 In:                       jah, 00:02:07-8      
4 In:                     [ja; ] 00:03:25-5      
5                     also jA:h; 00:03:16-6 (1.1)
6 Bz:               [E::hm;    ] 00:03:51-4 (3.0)
7 <NA>                      <NA>       <NA>  <NA>
8 <NA>                      <NA>       <NA>  <NA>

How can the regex be refined so that I get this desired output:

  Role                 Utterance  Timestamp   Gap
1 In:                   ja COOL; 00:04:24-6      
2      in den vier, FÜNF wochen, 00:04:57-8      
3 In:                       jah, 00:02:07-8      
4 In:                     [ja; ] 00:03:25-5      
5                     also jA:h; 00:03:16-6 (1.1)
6 Bz:               [E::hm;    ] 00:03:51-4 (3.0)
7 Bz:              [mhmh,      ]
8          in den bilLIE da war;

CodePudding user response：

An alternative to a complex regex is to use multiple extracts with simpler regexes. Afterwards convert any NA's to "" and strip unwanted whitespace.

library(dplyr)
library(tidyr)

data.frame(tst) %>%
  extract(tst, "Gap", "(\\(.*?\\))", remove = FALSE) %>%
  extract(tst, "Timestamp", "(#.*?#)", remove = FALSE) %>%
  extract(tst, c("Role", "Utterance"), "^(\\S :|)([^#]*)") %>%
  mutate(across(, coalesce, ""), Utterance = trimws(Utterance))

giving:

  Role                 Utterance    Timestamp   Gap
1  In:                  ja COOL; #00:04:24-6#      
2      in den vier, FÜNF wochen, #00:04:57-8#      
3  In:                      jah, #00:02:07-8#      
4  In:                    [ja; ] #00:03:25-5#      
5                     also jA:h; #00:03:16-6# (1.1)
6  Bz:              [E::hm;    ] #00:03:51-4# (3.0)
7  Bz:             [mhmh,      ]                   
8          in den bilLIE da war;

CodePudding user response：

You could update your pattern to use your 4 capture groups, and make the last part optional by optionally matching the 3rd group and then the 4th group and assert the end of the string:

library(tidyr)

tst <- c("In: ja COOL;  #00:04:24-6#  ",           
         "  in den vier, FÜNF wochen, #00:04:57-8# ",
         "In: jah,  #00:02:07-8# ",
         "In:     [ja; ] #00:03:25-5# [ja; ] #00:03:26-1#",
         "    also jA:h; #00:03:16-6# (1.1)",
         "Bz:        [E::hm;    ]  #00:03:51-4#  (3.0)  ",
         "Bz:    [mhmh,      ]",
         "  in den bilLIE da war;")     

data.frame(tst) %>%
  extract(col = tst,
          into = c("Role", "Utterance", "Timestamp", "Gap"),
          regex = "^(\\w{2}:\\s|\\s )([\\s\\S]*?)(?:\\s*#([^#] )(?:#\\s*(\\([0-9.] \\))?\\s*)?)?$")

Output

  Role                      Utterance  Timestamp   Gap
1 In:                        ja COOL; 00:04:24-6      
2           in den vier, FÜNF wochen, 00:04:57-8      
3 In:                            jah, 00:02:07-8      
4 In:      [ja; ] #00:03:25-5# [ja; ] 00:03:26-1      
5                          also jA:h; 00:03:16-6 (1.1)
6 Bz:                    [E::hm;    ] 00:03:51-4 (3.0)
7 Bz:                   [mhmh,      ]                 
8               in den bilLIE da war;