Reproduce “KPDrop: Improving Absent Keyphrase Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jishnu Ray Chowdhury, Seoyeon Park, Tuhin Kundu, Cornelia Caragea
KPDROP is a data augmentation and regularization technique that enhances the ability of keyphrase generation models to infer and produce absent keyphrases. The method consistently improved absent keyphrase F1 scores and recall across various models and datasets, and enabled substantial performance in low-resource settings by forcing models to leverage contextual information more deeply.
Key insights
- Artificially 'dropping' present keyphrases forces models to develop deeper contextual understanding and inferential capabilities, leading to better generation of genuinely absent keyphrases.
- Strategic data augmentation, such as KPDROP, can effectively address specific performance bottlenecks by creating challenging training examples that improve model robustness.
- The model-agnostic nature of KPDROP allows for easy integration with diverse existing and future keyphrase generation architectures without requiring architectural changes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.