The Naughtyformer: A Transformer Understands Offensive Humor
This work addresses the nuanced task of humor subtype classification, which is incremental as it builds on existing humor detection research.
The paper tackled the problem of detecting humor subtypes, particularly offensive humor, by introducing a new dataset from Reddit and using a finetuned Transformer model called Naughtyformer, achieving significantly better performance in offensiveness detection compared to state-of-the-art methods.
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.