metadata
language:
- zh
size_categories:
- 100K<n<1M
CFunSet: A Comprehensive Chinese Humor-Related Dataset
CFunSet is a comprehensive Chinese humor-related multi-task dataset designed to train CFunModel: A "Funny" Language Model Capable of Chinese Humor Generation and Processing. This dataset aggregates multiple existing datasets and includes diverse tasks such as humor generation, joke completion, humor classification, crosstalk response selection, and humor explanation.
📚 Dataset Overview
CFunSet consists of over 160,000 high-quality humor-related samples collected and curated from multiple sources. Each task is designed to assess and enhance various aspects of humor understanding and generation in Chinese.
🔥 Key Features
- Diverse Tasks: Includes joke generation, humor recognition, crosstalk response selection, and more.
- Multi-Source Aggregation: Combines data from Tieba-JokeBar, CrossDial, Chumor, HumorWB, Chinese-Humor-Sentiment, CCL-2019-Chinese-Humor-Sentiment.
📊 Data Structure
Each data sample follows the format:
"instruction": "Task-specific instruction describing the objective.",
"input": "Input text or prompt related to the task.",
"output": "Corresponding expected output or label."
📊 Dataset Statistics
| Source | Samples Count |
|---|---|
| Tieba-JokeBar | 18,41 |
| CrossDial-Dougen | 75,944 |
| CrossDial-Penggen | 43,372 |
| Chumor | 1,452 |
| HumorWB | 4,200 |
| Chinese-Humor-Sentiment | 4,860 |
| CCL-2019 Task 1 & Task 2 | 13,012 |
| Total | 164,200+ |
🤝 Citation
If you use CFunModel in your research or applications, please cite:
@misc{yu2025cfunmodelfunnylanguagemodel,
title={CFunModel: A "Funny" Language Model Capable of Chinese Humor Generation and Processing},
author={Zhenghan Yu and Xinyu Hu and Xiaojun Wan},
year={2025},
eprint={2503.20417},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2503.20417}, }
🎉 Happy Experimenting with CFunSet! 🎉