De-identifying Transcripts with ghosted

library(ghosted)

The ghosted package provides one user-facing function: ghost().

ghost() opens a local Shiny app for de-identifying batches of .vtt, .docx, and .txt transcript files. The app asks for the information needed to redact names and other terms, scans the selected input directory, lets you review likely names detected with local rule-based matching, and writes redacted transcripts to your selected output directory.

The app runs locally on your computer and does not use AI or remote services.

Start The App

ghost()

App Workflow

1. Enter the input directory containing transcript files.
2. Enter an output directory, or leave it blank to use the input directory.
3. Choose processing options:
   • output format: keep original formats, DOCX, TXT, or VTT
   • output filename suffix
   • redaction token
   • whether to add blank lines between turns
   • whether to redact interviewer names in transcript body text
   • whether to include common-name redaction
   • whether to print a console report
   • whether to show the completion notice
4. List known interviewer names, participant names, and any other terms to redact.
5. Click Scan directory.
6. Review likely names and classify them as interviewer, participant, or other.
7. Click Redact transcripts.

Re-running The App

When you run ghost() again in the same R session, the app preloads the directories, options, names, and likely-name selections from the previous run. If you are working on a different batch of transcripts, update the relevant sections before scanning and redacting.

Output Behavior

• Speaker labels at the start of a line or paragraph are normalized to Interviewer or Participant.
• Names elsewhere in transcript body text are replaced with the selected redaction token.
• Other terms are redacted exactly as entered.
• For DOCX/TXT to VTT output, cues are written without timestamps.
• Install officer for DOCX reads and writes.