DemMA-Dialog / README.md
hulehule's picture
Require email and manual review for dataset access
e02e033 verified
|
Raw History Blame Contribute Delete
3.67 kB
---
tags:
- demma
- dementia
- dialogue
- patient-simulation
- acl-2026
extra_gated_heading: "Request access to DemMA-Dialog"
extra_gated_description: >-
**Enter your email address below to request access to the dataset.**
Each request is reviewed manually. Access will be granted once your request is approved.
extra_gated_button_content: "Submit access request"
extra_gated_fields:
"Email address": text
"I have read the citation guidance": checkbox
extra_gated_prompt: |
## Citation
If you use this dataset in your research, please cite
[DemMA (Findings of ACL 2026)](https://aclanthology.org/2026.findings-acl.605/).
---
<h1 align="center">💬 DemMA-Dialog</h1>
<p align="center">
<strong>The context, reasoning, and behavior behind each conversation.</strong><br>
Dialogue corpus · Findings of ACL 2026
</p>
<p align="center">
<a href="https://aclanthology.org/2026.findings-acl.605/">📄 Paper</a> ·
<a href="https://aclanthology.org/2026.findings-acl.605.pdf">PDF</a> ·
<a href="https://github.com/RainieLLM/DemMA">💻 Code</a> ·
<a href="https://huggingface.co/hulehule/DemMA-Planner-SFT">🧠 Planner Model</a>
</p>
**Follow the reasoning behind each turn.** DemMA-Dialog brings patient profiles, conversation context, planner rationales, responses, and behavioral annotations together in JSONL records. It accompanies **DemMA**, an expert-guided language agent for multi-turn dementia patient simulation.
[![DemMA framework showing how patient profiles become annotated dialogue records](https://raw.githubusercontent.com/RainieLLM/DemMA/main/assets/demma_model.png)](https://raw.githubusercontent.com/RainieLLM/DemMA/main/assets/demma_model.png)
*Patient profiles → annotated conversations → a trained dialogue agent.*
## 🗂️ Inside a dialogue record
| Layer | What you can explore |
| --- | --- |
| 🧩 **Patient profile** | Persona, personality traits, and memory status. |
| 💬 **Conversation** | Dialogue history and the caregiver's current utterance. |
| 🧠 **Planner rationale** | Dialogue analysis, caregiver assessment, memory access, emotion prediction, and action planning. |
| 🎭 **Response & behavior** | The patient utterance, movement, facial-expression, and voice labels. |
The corpus is stored in [`DemMa_planner_labeled_dialogue_corpus.jsonl`](https://huggingface.co/datasets/hulehule/DemMA-Dialog/blob/main/DemMa_planner_labeled_dialogue_corpus.jsonl). Explore the records in the dataset viewer above, then see the [paper](https://aclanthology.org/2026.findings-acl.605/) for the methodology.
## 🔗 From data to model
Continue with the [DemMA Planner SFT checkpoint](https://huggingface.co/hulehule/DemMA-Planner-SFT) and the [official implementation](https://github.com/RainieLLM/DemMA) for data generation, training, and inference.
## 📚 Citation
If you find this dataset useful in your research, please consider citing our [DemMA paper](https://aclanthology.org/2026.findings-acl.605/).
<details>
<summary>Show BibTeX</summary>
```bibtex
@inproceedings{song-etal-2026-demma,
title = "{D}em{MA}: Dementia Multi-Turn Dialogue Agent with Expert-Guided Reasoning and Action Simulation",
author = "Song, Yutong and Wu, Jiang and Sharif, Kazi Shaharair and Zhang, Pengfei and Huang, Wenjun and Xu, Honghui and Dutt, Nikil and Rahmani, Amir M.",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
year = "2026",
publisher = "Association for Computational Linguistics",
pages = "12439--12470",
doi = "10.18653/v1/2026.findings-acl.605",
url = "https://aclanthology.org/2026.findings-acl.605/"
}
```
</details>