Persona-Chat

Dataset Summary

Persona-Chat is a high-quality multi-turn dialogue dataset designed to train conversational AI systems with consistent personality and style. Each participant in the dataset is assigned a persona—a short description or set of traits—which guides their responses throughout the conversation. This dataset enables AI models to learn to maintain coherent personas across dialogue turns and produce responses that reflect consistent characteristics, improving naturalness and engagement in human-like conversation.

The dataset contains real human conversations and captures both persona-driven and context-aware interactions. It is commonly used for research in dialogue modeling, persona-based chatbots, and human-AI interaction studies.

---

Data Structure

Fields

Field	Description
personality	List of persona traits describing a participant
utterances	Multi-turn conversation data
├── candidates	Candidate responses for the given turn
└── history	Past dialogue turns preceding the current turn

---

Splits

Split	Examples	Size (bytes)	Description
Train	16046	175333427	Used for model training
Validation	2832	30803403	Used for validation and tuning

Total dataset size: 206136830 bytes
Total number of dialogues: 18878

---

Usage Example

from datasets import load_dataset

ds = load_dataset("anezatra/persona-chat", split="train")
print(ds[0])

References

Zhang, S., Dinan, E., Urbanek, J., Szlam, A., Kiela, D., & Weston, J. (2018). Personalizing dialogue agents: I have a dog, do you have pets too?. arXiv preprint arXiv:1801.07243. (https://arxiv.org/abs/1801.07243)