Investigating Societal Biases in a Poetry Composition System
This work addresses biases in creative language applications, which are important for user interaction, but it is incremental as it builds on existing bias mitigation methods in a new domain.
The study tackled societal biases in a poetry composition system by investigating a pipeline to mitigate biases in next verse suggestions, finding that data augmentation via sentiment style transfer shows potential for this purpose.
There is a growing collection of work analyzing and mitigating societal biases in language understanding, generation, and retrieval tasks, though examining biases in creative tasks remains underexplored. Creative language applications are meant for direct interaction with users, so it is important to quantify and mitigate societal biases in these applications. We introduce a novel study on a pipeline to mitigate societal biases when retrieving next verse suggestions in a poetry composition system. Our results suggest that data augmentation through sentiment style transfer has potential for mitigating societal biases.