TalkTrace AI base
Copyright 2024-2026 Jami Schorling, Dennis Hauk, Simon Filler

This product includes software originally developed by Jami Schorling and
Dennis Hauk at the Chair for Research on Teaching and Learning in Civic
Education, Leipzig University, as TalkTrace AI
(https://github.com/talktrace-ai/talktrace-ai).

The base distribution is maintained by Simon Filler at TU Dortmund University
(https://idif.sowi.tu-dortmund.de/institut/) and is a reduced, AGPL-3.0
licensed sibling of an actively developed internal research fork.

Original authors:
  - Jami Schorling   (ORCID: 0009-0005-9007-2896)
  - Dennis Hauk      (ORCID: 0000-0002-5779-2876)

Maintainer of the base distribution:
  - Simon Filler     (ORCID: 0009-0008-8736-8831)


Licensing history
-----------------

The upstream repository (talktrace-ai/talktrace-ai) carries a Creative Commons
Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) notice. The base
distribution is released under the GNU Affero General Public License,
Version 3.0 (AGPL-3.0), with the explicit consent of the original authors
Jami Schorling and Dennis Hauk to relicense the derived work under a
different OSI-approved license.

Consent was given on 2026-05-26.

(Maintainer: keep the underlying written consent — email, signed message,
or equivalent — on file with the project records in case the licensing
chain is ever questioned.)


Optional audio-transcription module (noScribe)
----------------------------------------------

TalkTrace AI base ships an optional local audio-transcription feature (the
"Transcription" tab). It does NOT bundle, link against, import, or
redistribute any transcription engine. Instead, at the user's explicit
request, it downloads the standalone open-source project noScribe and
invokes it strictly as a separate subprocess:

  noScribe — Copyright (c) Kai Dröge
  https://github.com/kaixxx/noScribe
  Licensed under the GNU General Public License, Version 3.0 (GPL-3.0).

Because noScribe runs only as an independent program at arm's length (no
shared address space, no linking, no shared code), it is not a derivative
work of TalkTrace AI base and TalkTrace AI base is not a derivative work of
noScribe. The two licenses (AGPL-3.0 for TalkTrace, GPL-3.0 for noScribe)
therefore apply independently to their respective codebases.

The engine is installed on demand into a per-user directory
(%LOCALAPPDATA%\TalkTraceAI\noscribe-engine on Windows) using the `uv`
package manager, and pulls these third-party components from their public
upstream sources:

  - faster-whisper (CTranslate2) Whisper models:
      "fast"    — mukowaty/faster-whisper-int8
                  (faster-whisper-large-v3-turbo-int8)
      "precise" — mobiuslabsgmbh/faster-whisper-large-v3-turbo
  - pyannote.audio speaker-diarization models (bundled with noScribe,
      CC-BY-4.0)
  - PyTorch, torchaudio, torchcodec, pyannote.audio, faster-whisper and
      their transitive dependencies, each under its own license.

All transcription runs entirely on the user's machine; audio data is not
transmitted to any maintainer-controlled or third-party server.


Built-in T-SEDA codebook template
----------------------------------

TalkTrace AI base ships an optional built-in codebook template for the
analysis of dialogic classroom talk, based on:

  T-SEDA Collective (2023). Toolkit for Systematic Educational Dialogue
  Analysis (T-SEDA): A resource for inquiry into practice. v.9.
  University of Cambridge. https://camtree.learnworlds.com/t-seda
  Released under a Creative Commons Attribution licence (CC BY).
  German code abbreviations follow the official German edition of the
  toolkit ("Toolkit für die systematische Analyse dialogischer
  Unterrichtsgespräche", translation supervised by E. Calcagni).

T-SEDA explicitly invites local adaptation. The template in
talktrace_ai/examples/tseda.py keeps the official code logic and labels,
condenses the definitions for LLM-prompt use, and replaces the original
examples with subject-neutral classroom anchors. It is an adaptation, not
a verbatim reproduction; any coding errors introduced by the adaptation
are the responsibility of this project, not of the T-SEDA Collective.


Consent-declaration generator
------------------------------

TalkTrace AI base ships an optional consent-declaration generator (the
"Consent" tab) that produces a print-ready GDPR Art. 13 consent form for the
training context. The wording is adapted from:

  Consent-Gen-RDMO — Bernd Zey and Wibke Kleina,
  Research Data Management group, TU Dortmund University
  https://github.com/berndzey/Consent-Gen-RDMO
  Dedicated to the public domain under CC0 1.0 Universal.

Because the upstream project is released under CC0 1.0, its text may be
reused and adapted without restriction. The generated declaration is an aid,
not legal advice; the generator surfaces this disclaimer in the form and in
the document footer. Each deployment must have it reviewed by its own data
protection officer and adapt it to local/regional legal requirements.
