COBRANext.js LogoFrames
Contextual Reasoning about Effects and Harms of Offensive Statements

Paper ->

Find in-depth information about Cobra Frames.

Code ->

Use the Cobra Frames codebase to build your own.

Demo ->

Explore the Cobra Frames demo.

Data ->

Explore and use the Cobra Frames data.

A quick glance

Introducing COBRA frames, a context-aware approach to understanding offensive language. It emphasizes the significance of social and situational context in interpreting statements. We also created COBRACORPUS, a dataset of offensive statements paired with context and explanations. We trained models to generate COBRA explanations, finding that context-aware models significantly outperform context-agnostic ones, especially when the context changes the offensiveness of the statement. Our work underscores the importance of considering social factors in NLP models.

👉 (on the right), as illstrated in the figure, our formalism enables counterfactual reasoning about contexts that invert the statements’ offensiveness.

Next.js Logo

Citations

@inproceedings{zhou2023cobraframes,
  title={{COBRA Frames: Contextual Reasoning about 
Effects and Harms of Offensive Statements}},
  author={Zhou, Xuhui and Zhu, Hao and Yerukola, Akhila and Davidson, Thomas and
  Hwang, Jena D. and Swayamdipta, Swabha and Sap, Maarten},
  year={2023},
  booktitle={{Findings of ACL}},
  url={{http://arxiv.org/abs/2306.01985}}
}