obsidian-plugin/core-hn_pseudobsidian-ization
1
0
Fork 0
mirror of https://github.com/core-hn/pseudobsidian-ization.git synced 2026-07-22 06:57:06 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

docs: add noScribe integration & audio pseudonymization to Phase 10 roadmap

Browse source 
- VttParser with word-level timestamps and speaker diarization
- Timestamp adjustment UI with ±1s audio preview (Web Audio API)
- Audio bleep export (WAV, pseudonymized segments replaced by sine tone)
- Speaker-aware pseudonymization rules
This commit is contained in:
Axelle Abbadie 2026-05-16 02:13:00 +02:00
parent 3d4ead5a3b
commit 5a9e1f0581
1 changed files with 10 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

11
ROADMAP.md
View file

@ -227,7 +227,16 @@ Goal: consolidate all existing features and add the EMCA-specific functions that
- [ ] [Meld Encrypt](https://github.com/meld-cp/obsidian-encrypt) integration in Exports tab for encrypting correspondence tables and pseudonymized exports
- [ ] Resolve remaining open questions (SPECS §20)
**Testable (v0.2.0):** stable end-to-end workflow on a real corpus of 10 interviews.
### noScribe integration & audio pseudonymization
[noScribe](https://github.com/kaixxx/noScribe) is a local transcription tool (Whisper + pyannote) widely used in qualitative research. It produces VTT files with **word-level timestamps** and **speaker diarization**.
- [ ] **`VttParser.ts`** — import noScribe VTT: word-level timestamps, speaker labels, round-trip guarantee; auto-conversion to `.md` on import alongside `.mapping.json`
- [ ] **Timestamp adjustment UI** — when a word timestamp from Whisper is imprecise, the researcher can fine-tune it: playback of ±1 s around the word (Web Audio API, no external dependency), editable start/end fields, saved back to the mapping metadata
- [ ] **Audio bleep export** — once terms are pseudonymized and timestamps are validated, generate an audio export where each pseudonymized occurrence is replaced by a sine tone at the exact word position; export as WAV via Web Audio API
- [ ] **Speaker-aware pseudonymization** — rules can be scoped to a specific speaker label (e.g. `SPEAKER_1`); when a name is pseudonymized in one turn, offer to apply the rule to all turns by the same speaker
**Testable (v0.2.0):** stable end-to-end workflow on a real corpus of 10 interviews, including noScribe VTT import, audio bleep export, and timestamp adjustment.
---