leonardtang / the-naughtyformer

The Naughtyformer: A Transformer Understands Offensive Humor (AAAI 2023)

The Naughtyformer: A Transformer Understands Offensive Humor

Jokes are intentionally written to be funny, but not all jokes are created the same. Some jokes may be fit for a classroom of kindergarteners, but others are best reserved for a more mature audience. While recent work has shown impressive results on humor detection in text, here we instead investigate the more nuanced task of detecting humor subtypes, especially of the less innocent variety. To that end, we introduce a novel jokes dataset filtered from Reddit and solve the subtype classification task using a finetuned Transformer dubbed the Naughtyformer. Moreover, we show that our model is significantly better at detecting offensiveness in jokes compared to state-of-the-art methods.

Dataset

To train the Naughtyformer, we introduce a dataset of 92,153 total jokes across categories of 1) Clean Jokes, 2) Dark Jokes, and 3) Dirty Jokes. We also include a fourth category, News, representing a non-joke.

The cleaned and post-processed data is found in cleaned_data.csv. The original Reddit is found in raw_data/.

Model Checkpoint

Our DeBERTa checkpoint for the Naughtyformer can be found here.

Citation

If you find this useful in your research, please consider citing:

@article{tang2022naughtyformer,
  title={The Naughtyformer: A Transformer Understands Offensive Humor},
  author={Tang, Leonard and Cai, Alexander and Li, Steve and Wang, Jason},
  journal={arXiv preprint arXiv:2211.14369},
  year={2022}
}

the-naughtyformer's Issues

What is the input syntax?

Did you give f"{title} {sep_token} {self_text}"? Or some other variation?

The model files are not correct

The optimizer and pytorch_mode.bin are both 22M where the original microsoft/deberta-base is 559M. I think this didnt get fully loaded.

The model files are deleted

It looks like you deleted the model files: b0f0b34