UKPLab/Graph2Counsel

Dataset Card

Paper: Graph2Counsel: Clinically Grounded Synthetic Counseling Dialogue Generation from Client Psychological Graphs

Language(s) (NLP): English

license: cc-by-sa-4.0

Dataset Summary

Graph2Counsel is a synthetic counseling session dataset generated from Client Psychological Graphs (CPGs). The dataset provides the CPG, generated diverse client profiles and dialogues from this CPG as well as counselor strategies corresponding to each CPG collected from the real counseling sessions. Prior synthetic counseling datasets often rely on unstructured or semi-structured text inputs and overlook structural dependencies between a client's cognitive, emotional, and behavioral states, often producing psychologically inconsistent interactions and reducing data realism and quality. So, we generate the Graph2Counsel dataset using a framework for generating synthetic counseling sessions grounded in Client Psychological Graphs (CPGs) that encode relationships among clients’ thoughts, emotions, and behaviors. This datasets shows improved expert evaluation performance over existing synthetic counseling datasets on specificity, counselor competence, authenticity, conversational flow, and safety.

[image] From a real therapy transcript, we extract a Client Psychological Graph (CPG) (using the method from [Ong et al., 2025](https://arxiv.org/abs/2512.05836)), whose nodes represent psychological processes (e.g., fear of judgment) and edges capture their relationships (e.g., excites or inhibits). From the same transcript, we also extract counselor strategies (e.g., reframing, empathy building). We then use the CPG to generate diverse client profiles, and finally combine the profile, CPG, and counselor strategies to generate synthetic counseling dialogues.

Dataset Structure

Each record contains the following fields:

Data Sources

• —CPGs: Derived from real anonymised real counseling sessions. Each CPG encodes structural dependencies between a client's cognitive, emotional, and behavioral states
• —Client profiles and session dialogues: Synthetically generated from the CPGs. Multiple variations (variation index) are produced per CPG to capture diverse but plausible client presentations and therapeutic trajectories.
• —Counselor strategies: Curated per session, reflecting the therapeutic approaches used by the real counselor throughout the dialogue.

Intended Use

This dataset is intended for:

• —Fine-tuning LLMs for mental health counseling dialogue generation
• —Training models to ground therapeutic responses in structured clinical knowledge (CPGs)
• —Studying the alignment between client psychological graphs, counselor strategy, and dialogue quality

Usage

from datasets import load_dataset

ds = load_dataset("UKPLab/Graph2Counsel")

Citation

If you use this dataset in your research, please cite:

@misc{mandal2026graph2counselclinicallygroundedsynthetic,
      title={Graph2Counsel: Clinically Grounded Synthetic Counseling Dialogue Generation from Client Psychological Graphs}, 
      author={Aishik Mandal and Hiba Arnaout and Clarissa W. Ong and Juliet Bockhorst and Kate Sheehan and Rachael Moldow and Tanmoy Chakraborty and Iryna Gurevych},
      year={2026},
      eprint={2604.20382},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2604.20382}, 
}

Contact

For questions or feedback regarding this dataset, please contact: aishik.mandal@tu-darmstadt.de

Ethical Considerations

• —All client profiles and session dialogues are synthetically generated and do not correspond to real individuals.
• —The underlying CPGs are derived from real counseling sessions and have been anonymised.
• —This dataset is intended strictly for research purposes. It should not be used to deploy autonomous mental health interventions without appropriate clinical oversight.
• —Models fine-tuned on this dataset should be evaluated carefully before any clinical application.