Annotated corpus of 59 simulated German teacher-parent counseling conversations (5,937 phase annotations, 904 technique annotations across ~6k sentences) from the paper:
Annotating Conversational Phases and Communication Techniques: A Corpus of German Teacher-Parent Counseling Conversations
Tobias Hallmen, Kathrin Gietl, Karoline Hillesheim, Annemarie Friedrich, Elisabeth André
University of Augsburg
LREC 2026
| File/Folder | Description |
|---|---|
data/gold_standard.csv |
Gold standard annotations (59 conversations, English labels) |
data/transcripts/ |
59 transcript files (German, sentence-segmented) |
guidelines/ |
Annotation guidelines (German) |
code/bert.py |
BERT-based baseline (fine-tuning, 4-fold CV) |
code/llm_eval.py |
LLM zero-shot baseline (requires Ollama) |
| Column | Description |
|---|---|
dataset |
Semester index (0–3), used as cross-validation fold |
conversation_id |
Zero-padded document identifier (e.g. 053, 037) |
sentence_id |
Sentence index within the conversation |
type |
Phase or Technique |
sentence |
Sentence text (T: = teacher, P: = parent, simul. = simultaneous speech) |
label |
Gold standard label (see below) |
sublabel |
Sub-category label |
Beginning, Informational, Argumentative, Decision-Making, Concluding
Paraphrasing, Structuring, Verbalizing
(sentences without a technique annotation have label Other in model training)
import pandas as pd
df = pd.read_csv('data/gold_standard.csv')
# Phase annotations for one conversation
phases = df.query('conversation_id == "053" and type == "Phase"')
# Technique gold standard (4-way classification incl. Other)
techniques = df.query('type == "Technique"')BERT fine-tuning (requires GPU, HuggingFace transformers):
pip install transformers datasets scikit-learn
python code/bert.pyLLM zero-shot (requires Ollama running locally):
ollama pull gemma3:4b
python code/llm_eval.py| Split | Conversations | Phase annotations | Technique annotations |
|---|---|---|---|
| Semester 1 (dataset=0) | 6 | 820 | 37 |
| Semester 2 (dataset=1) | 20 | 2,154 | 444 |
| Semester 3 (dataset=2) | 17 | 1,617 | 270 |
| Semester 4 (dataset=3) | 16 | 1,346 | 153 |
| Total | 59 | 5,937 | 904 |
This dataset is released under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) license.
@inproceedings{hallmen-etal-2026-annotating,
title = {Annotating Conversational Phases and Communication Techniques: A Corpus of German Teacher-Parent Counseling Conversations},
author = {Hallmen, Tobias and Gietl, Kathrin and Hillesheim, Karoline and Friedrich, Annemarie and André, Elisabeth},
booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference (LREC 2026)},
month = {May},
year = {2026},
pages = {6361--6372},
address = {Palma, Mallorca, Spain},
publisher = {European Language Resources Association (ELRA)},
editor = {Piperidis, Stelios and Bel, Núria and van den Heuvel, Henk and Ide, Nancy and Krek, Simon and Toral, Antonio},
doi = {10.63317/4ukmi5gkdygr},
abstract = {Teacher-parent conversations are critical for student success, yet teachers often lack structured training in counseling communication skills. We present the first annotated corpus of teacher-parent counseling conversations consisting of 59 German dialogues (approximately 6k sentences, 21k annotations) simulated by prospective elementary school teachers, peers, and professional actors. The corpus features theory-grounded annotations for conversational phases (Beginning, Informational, Argumentative, Decision-Making, Concluding) and communication techniques (Paraphrasing, Verbalizing, Structuring). We provide detailed annotation guidelines operationalizing established counseling pedagogy frameworks for computational analysis. Inter-annotator agreement analysis reveals substantial agreement (Fleiss' k = 0.669 to 0.724, Krippendorff's a = 0.666 to 0.735). Our analysis reveals confusion patterns, providing insights into counseling discourse structure. Baseline experiments with BERT-based models and open-source LLMs achieve F1 scores of up to 71% depending on task and model. The corpus, guidelines, and baseline code are publicly available under CC BY-NC-SA 4.0 license, enabling research on automated dialogue analysis and AI-based training tools for teacher education.},
}We thank all students who participated in the seminar simulations and annotation work. This work was partially funded by the Stiftung Innovation in der Hochschullehre (KodiLL project, FBM2020, grant no. 120).