Header

UZH-Logo

Maintenance Infos

Summary Refinement through Denoising


Nikolov, Nikola I; Calmanovici, Alessandro; Hahnloser, Richard H R (2019). Summary Refinement through Denoising. In: International Conference on Recent Advances in Natural Language Processing (RANLP 2019), Varna, Bulgaria, 31 August 2019 - 6 September 2019, RANLP.

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.

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.

Statistics

Downloads

12 downloads since deposited on 14 Feb 2020
2 downloads since 12 months
Detailed statistics

Additional indexing

Item Type:Conference or Workshop Item (Paper), refereed, original work
Communities & Collections:07 Faculty of Science > Institute of Neuroinformatics
Dewey Decimal Classification:570 Life sciences; biology
Scopus Subject Areas:Physical Sciences > Software
Physical Sciences > Computer Science Applications
Physical Sciences > Artificial Intelligence
Physical Sciences > Electrical and Electronic Engineering
Language:English
Event End Date:6 September 2019
Deposited On:14 Feb 2020 10:33
Last Modified:16 Jun 2022 07:05
Publisher:RANLP
OA Status:Green
Free access at:Official URL. An embargo period may apply.
Official URL:https://acl-bg.org/proceedings/2019/RANLP%202019/pdf/RANLP097.pdf
  • Content: Published Version