metadata
license: apache-2.0
Stephen Dataset
The Stephen Dataset is a humor-infused dialogue and joke collection designed to train conversational AI with a sarcastic, witty, and opinionated personality — perfect for code-roasting and banter.
Dataset Summary
This dataset combines:
- Friends TV show transcripts (for Chandler Bing–style sarcasm)
- Reddit jokes & programming humor
- Sarcasm headlines
- Developer-themed one-liners
It is intended for training Stephen, a conversational coding assistant with sharp wit and self-aware humor.
Supported Tasks
- Conversational AI
- Sarcasm generation
- Programming humor generation
- Code review banter
Languages
- English
Dataset Structure
Format: CSV with two columns:
character— The speaker (e.g., Chandler, Stephen, Narrator)line— The sarcastic or humorous line of dialogue
Example:
| character | line |
|---|---|
| Stephen | Oh, you broke production? Great! Now we can finally test our incident response plan. |
Licensing
This dataset is released under the Apache 2.0 License.
Citation
@misc{stephen_dataset,
title = {Stephen Dataset: Sarcastic Humor for Conversational AI},
author = {dgtalbug},
year = {2025},
howpublished = {\url{https://huggingface.co/datasets/dgtalbug/stephen-dataset}}
}