Fair Summarization: Bridging Quality and Diversity in Extractive Summaries

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Nezhad, Sina Bagheri, Bandyapadhyay, Sayan, Agrawal, Ameeta
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910879654608896
author Nezhad, Sina Bagheri
Bandyapadhyay, Sayan
Agrawal, Ameeta
author_facet Nezhad, Sina Bagheri
Bandyapadhyay, Sayan
Agrawal, Ameeta
contents Fairness in multi-document summarization of user-generated content remains a critical challenge in natural language processing (NLP). Existing summarization methods often fail to ensure equitable representation across different social groups, leading to biased outputs. In this paper, we introduce two novel methods for fair extractive summarization: FairExtract, a clustering-based approach, and FairGPT, which leverages GPT-3.5-turbo with fairness constraints. We evaluate these methods using Divsumm summarization dataset of White-aligned, Hispanic, and African-American dialect tweets and compare them against relevant baselines. The results obtained using a comprehensive set of summarization quality metrics such as SUPERT, BLANC, SummaQA, BARTScore, and UniEval, as well as a fairness metric F, demonstrate that FairExtract and FairGPT achieve superior fairness while maintaining competitive summarization quality. Additionally, we introduce composite metrics (e.g., SUPERT+F, BLANC+F) that integrate quality and fairness into a single evaluation framework, offering a more nuanced understanding of the trade-offs between these objectives. Our code is available online.
format Preprint
id arxiv_https___arxiv_org_abs_2411_07521
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Fair Summarization: Bridging Quality and Diversity in Extractive Summaries
Nezhad, Sina Bagheri
Bandyapadhyay, Sayan
Agrawal, Ameeta
Computation and Language
Artificial Intelligence
Fairness in multi-document summarization of user-generated content remains a critical challenge in natural language processing (NLP). Existing summarization methods often fail to ensure equitable representation across different social groups, leading to biased outputs. In this paper, we introduce two novel methods for fair extractive summarization: FairExtract, a clustering-based approach, and FairGPT, which leverages GPT-3.5-turbo with fairness constraints. We evaluate these methods using Divsumm summarization dataset of White-aligned, Hispanic, and African-American dialect tweets and compare them against relevant baselines. The results obtained using a comprehensive set of summarization quality metrics such as SUPERT, BLANC, SummaQA, BARTScore, and UniEval, as well as a fairness metric F, demonstrate that FairExtract and FairGPT achieve superior fairness while maintaining competitive summarization quality. Additionally, we introduce composite metrics (e.g., SUPERT+F, BLANC+F) that integrate quality and fairness into a single evaluation framework, offering a more nuanced understanding of the trade-offs between these objectives. Our code is available online.
title Fair Summarization: Bridging Quality and Diversity in Extractive Summaries
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2411.07521