Question

I am using a Voice-to-Text application which gives transcription files as output.. The transcribed text contains a few tags like (s) (for sentence beginning)..(/s)( for sentence end ).. (VOCAL_NOISE)(for un-recognized words).. but the text also contains unwanted tags like (VOCAL_N) , (VOCAL_NOISED) , (VOCAL_SOUND), (UNKNOWN).. i am using SED to process the text.. but cannot write an appropriate regex to replace all other tags except (s), (/s) and (VOCAL_NOISE), with the tag ~NS.. would appreciate if someone could help me with it..

Example text:

(s) Hi Stacey , this is Stanley (/s) (s) I would (VOCAL_N) appreciate if you could call (UNKNOWN) and let him know I want an appointment (VOCAL_NOISE) with him (/s)

Output should be:

(s) Hi Stacey , this is Stanley (/s) (s) I would ~NS appreciate if you could call ~NS and let him know I want an appointment (VOCAL_NOISE) with him (/s)

Answer 1

This should take care of it:

sed 's|([^)]*)|\n&\n|g;s@\n\((/\?s)\|(VOCAL_NOISE)\)\n@\1@g;s|\n\(([^)]*)\)\n|~NS|g' inputfile

Explanation:

• s|([^)]*)|\n&\n|g – divide the line by putting every parenthesized string between two newlines
• s@\n\((/\?s)\|(VOCAL_NOISE)\)\n@\1@g – remove the newlines around “(s)”, “(/s)” and
  “(VOCAL_NOISE)” (keepers)
• s|\n\(([^)]*)\)\n|~NS|g – replace anything else between newlines that is within parentheses with “~NS”

This works since newlines are guaranteed not to appear within a newly read line of text.

Edit: Shortened the command by using alternation \(foo\|bar\)

Previous version:

    sed 's|([^)]*)|\n&\n|g;s|\n\((/\?s)\)\n|\1|g; s|\n\((VOCAL_NOISE)\)\n|\1|g;s|\n\(([^)]*)\)\n|~NS|g' inputfile