Summary of Getting Serious About Humor: Crafting Humor Datasets with Unfunny Large Language Models, by Zachary Horvitz et al.
Getting Serious about Humor: Crafting Humor Datasets with Unfunny Large Language Models
by Zachary Horvitz, Jingru Chen, Rahul Aditya, Harshvardhan Srivastava, Robert West, Zhou Yu, Kathleen McKeown
First submitted to arxiv on: 23 Feb 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper investigates the potential of large language models (LLMs) in generating synthetic data for humor detection. Specifically, it explores whether LLMs can “unfun” jokes by editing texts to make them non-humorous. The authors benchmark their approach on an existing human dataset and demonstrate that current LLMs, such as GPT-4, are capable of successfully generating unfunny versions of humorous texts. This ability is evaluated both subjectively by humans and objectively using the downstream task of humor detection. Furthermore, the paper extends its approach to a code-mixed English-Hindi humor dataset, where it finds that GPT-4’s synthetic data is highly rated by bilingual annotators and provides challenging adversarial examples for humor classifiers. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how computers can help us detect what’s funny or not. Right now, detecting humor is hard because we don’t have enough good examples of jokes with non-joke versions to compare them to. The authors want to see if big language models can create these “unfun” jokes themselves. They tested their idea on a dataset of human-labeled humorous texts and found that the computers were pretty good at making jokes not funny anymore! They even tried it out on jokes in two languages, English and Hindi, and people thought the computer-made unfunny jokes were really well done. This could help us create better ways to tell whether something is meant to be funny or not. |
Keywords
* Artificial intelligence * Gpt * Synthetic data
