NeuTral Rewriter: A Rule-Based and Neural Approach to Automatic Rewriting into Gender-Neutral Alternatives

Eva Vanmassenhove; Chris Emmery; Dimitar Shterionov

doi:10.18653/v1/2021.emnlp-main.704

NeuTral Rewriter: A Rule-Based and Neural Approach to Automatic Rewriting into Gender-Neutral Alternatives

Eva Vanmassenhove, Chris Emmery, Dimitar Shterionov

Cognitive Science & AI

Research output: Contribution to conference › Paper › Scientific › peer-review

43 Downloads (Pure)

Abstract

Recent years have seen an increasing need for gender-neutral and inclusive language. Within the field of NLP, there are various mono- and bilingual use cases where gender inclusive language is appropriate, if not preferred due to ambiguity or uncertainty in terms of the gender of referents. In this work, we present a rule-based and a neural approach to gender-neutral rewriting for English along with manually curated synthetic data (WinoBias+) and natural data (OpenSubtitles and Reddit) benchmarks. A detailed manual and automatic evaluation highlights how our NeuTral Rewriter, trained on data generated by the rule-based approach, obtains word error rates (WER) below 0.18% on synthetic, in-domain and out-domain test sets.

Original language	English
Pages	8940–8948
Number of pages	8
DOIs	https://doi.org/10.18653/v1/2021.emnlp-main.704
Publication status	Published - 13 Sept 2021
Event	The 2021 Conference on Empirical Methods in Natural Language Processing - Punta Cana, Punta Cana, Dominica Duration: 7 Nov 2022 → 11 Nov 2022 https://2021.emnlp.org/

Conference

Conference	The 2021 Conference on Empirical Methods in Natural Language Processing
Abbreviated title	EMNLP 2021
Country/Territory	Dominica
City	Punta Cana
Period	7/11/22 → 11/11/22
Internet address	https://2021.emnlp.org/

Keywords

cs.CL
cs.AI

Access to Document

10.18653/v1/2021.emnlp-main.704Licence: CC BY

2109.06105v1

Cite this

@conference{6e2d5f84f04f415fa0c00825f3eb856e,

title = "NeuTral Rewriter: A Rule-Based and Neural Approach to Automatic Rewriting into Gender-Neutral Alternatives",

abstract = " Recent years have seen an increasing need for gender-neutral and inclusive language. Within the field of NLP, there are various mono- and bilingual use cases where gender inclusive language is appropriate, if not preferred due to ambiguity or uncertainty in terms of the gender of referents. In this work, we present a rule-based and a neural approach to gender-neutral rewriting for English along with manually curated synthetic data (WinoBias+) and natural data (OpenSubtitles and Reddit) benchmarks. A detailed manual and automatic evaluation highlights how our NeuTral Rewriter, trained on data generated by the rule-based approach, obtains word error rates (WER) below 0.18% on synthetic, in-domain and out-domain test sets. ",

keywords = "cs.CL, cs.AI",

author = "Eva Vanmassenhove and Chris Emmery and Dimitar Shterionov",

year = "2021",

month = sep,

day = "13",

doi = "10.18653/v1/2021.emnlp-main.704",

language = "English",

pages = "8940–8948",

note = "The 2021 Conference on Empirical Methods in Natural Language Processing, EMNLP 2021 ; Conference date: 07-11-2022 Through 11-11-2022",

url = "https://2021.emnlp.org/",

}

NeuTral Rewriter: A Rule-Based and Neural Approach to Automatic Rewriting into Gender-Neutral Alternatives. / Vanmassenhove, Eva ; Emmery, Chris ; Shterionov, Dimitar.
2021. 8940–8948 Paper presented at The 2021 Conference on Empirical Methods in Natural Language Processing, Punta Cana, Dominica.

Research output: Contribution to conference › Paper › Scientific › peer-review

TY - CONF

T1 - NeuTral Rewriter

T2 - The 2021 Conference on Empirical Methods in Natural Language Processing

AU - Vanmassenhove, Eva

AU - Emmery, Chris

AU - Shterionov, Dimitar

PY - 2021/9/13

Y1 - 2021/9/13

N2 - Recent years have seen an increasing need for gender-neutral and inclusive language. Within the field of NLP, there are various mono- and bilingual use cases where gender inclusive language is appropriate, if not preferred due to ambiguity or uncertainty in terms of the gender of referents. In this work, we present a rule-based and a neural approach to gender-neutral rewriting for English along with manually curated synthetic data (WinoBias+) and natural data (OpenSubtitles and Reddit) benchmarks. A detailed manual and automatic evaluation highlights how our NeuTral Rewriter, trained on data generated by the rule-based approach, obtains word error rates (WER) below 0.18% on synthetic, in-domain and out-domain test sets.

AB - Recent years have seen an increasing need for gender-neutral and inclusive language. Within the field of NLP, there are various mono- and bilingual use cases where gender inclusive language is appropriate, if not preferred due to ambiguity or uncertainty in terms of the gender of referents. In this work, we present a rule-based and a neural approach to gender-neutral rewriting for English along with manually curated synthetic data (WinoBias+) and natural data (OpenSubtitles and Reddit) benchmarks. A detailed manual and automatic evaluation highlights how our NeuTral Rewriter, trained on data generated by the rule-based approach, obtains word error rates (WER) below 0.18% on synthetic, in-domain and out-domain test sets.

KW - cs.CL

KW - cs.AI

U2 - 10.18653/v1/2021.emnlp-main.704

DO - 10.18653/v1/2021.emnlp-main.704

M3 - Paper

SP - 8940

EP - 8948

Y2 - 7 November 2022 through 11 November 2022

ER -

NeuTral Rewriter: A Rule-Based and Neural Approach to Automatic Rewriting into Gender-Neutral Alternatives

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this