Abstract
We propose a simple method for post-processing the outputs of a text summarization system in order to refine its overall quality. Our approach is to train text-to-text rewriting models to correct information redundancy errors that may arise during summarization. We train on synthetically generated noisy summaries, testing three different types of noise that introduce out-of-context information within each summary. When applied on top of extractive and abstractive summarization baselines, our summary denoising models yield metric improvements while reducing redundancy.- Anthology ID:
- R19-1097
- Volume:
- Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2019)
- Month:
- September
- Year:
- 2019
- Address:
- Varna, Bulgaria
- Venue:
- RANLP
- SIG:
- Publisher:
- INCOMA Ltd.
- Note:
- Pages:
- 837–843
- Language:
- URL:
- https://aclanthology.org/R19-1097
- DOI:
- 10.26615/978-954-452-056-4_097
- Cite (ACL):
- Nikola I. Nikolov, Alessandro Calmanovici, and Richard Hahnloser. 2019. Summary Refinement through Denoising. In Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2019), pages 837–843, Varna, Bulgaria. INCOMA Ltd..
- Cite (Informal):
- Summary Refinement through Denoising (Nikolov et al., RANLP 2019)
- PDF:
- https://preview.aclanthology.org/auto-file-uploads/R19-1097.pdf
- Code
- ninikolov/summary-denoising