Understanding Gender Biases in Knowledge Base Embeddings

Yupei Du

doi:10.18653/v1/2022.acl-long.98

Understanding Gender Biases in Knowledge Base Embeddings

Yupei Du

Sub Natural Language Processing

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

Abstract

Knowledge base (KB) embeddings have been shown to contain gender biases. In this paper, we study two questions regarding these biases: how to quantify them, and how to trace their origins in KB? Specifically, first, we develop two novel bias measures respectively for a group of person entities and an individual person entity. Evidence of their validity is observed by comparison with real-world census data. Second, we use the influence function to inspect the contribution of each triple in KB to the overall group bias. To exemplify the potential applications of our study, we also present two strategies (by adding and removing KB triples) to mitigate gender biases in KB embeddings.

Original language	English
Title of host publication	Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Editors	S. Muresan, P. Nakov, A. Villavicencio
Publisher	Association for Computational Linguistics
Pages	1381–1395
DOIs	https://doi.org/10.18653/v1/2022.acl-long.98
Publication status	Published - 2022
Event	Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) - Dublin, Ireland Duration: 22 May 2022 → 27 May 2022

Conference

Conference	Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Country/Territory	Ireland
City	Dublin
Period	22/05/22 → 27/05/22

Access to Document

10.18653/v1/2022.acl-long.98Licence: CC BY

2022.acl-long.98Final published version, 1.1 MBLicence: CC BY

Cite this

@inproceedings{c97362a0b2ff40a9b8d008ca78f5e751,

title = "Understanding Gender Biases in Knowledge Base Embeddings",

abstract = "Knowledge base (KB) embeddings have been shown to contain gender biases. In this paper, we study two questions regarding these biases: how to quantify them, and how to trace their origins in KB? Specifically, first, we develop two novel bias measures respectively for a group of person entities and an individual person entity. Evidence of their validity is observed by comparison with real-world census data. Second, we use the influence function to inspect the contribution of each triple in KB to the overall group bias. To exemplify the potential applications of our study, we also present two strategies (by adding and removing KB triples) to mitigate gender biases in KB embeddings.",

author = "Yupei Du",

year = "2022",

doi = "10.18653/v1/2022.acl-long.98",

language = "English",

pages = "1381–1395",

editor = "S. Muresan and P. Nakov and A. Villavicencio",

booktitle = "Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",

publisher = "Association for Computational Linguistics",

address = "United States",

note = "Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) ; Conference date: 22-05-2022 Through 27-05-2022",

}

Du, Y 2022, Understanding Gender Biases in Knowledge Base Embeddings. in S Muresan, P Nakov & A Villavicencio (eds), Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Association for Computational Linguistics, pp. 1381–1395, Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Dublin, Ireland, 22/05/22. https://doi.org/10.18653/v1/2022.acl-long.98

Understanding Gender Biases in Knowledge Base Embeddings. / Du, Yupei.
Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). ed. / S. Muresan; P. Nakov; A. Villavicencio. Association for Computational Linguistics, 2022. p. 1381–1395.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - Understanding Gender Biases in Knowledge Base Embeddings

AU - Du, Yupei

PY - 2022

Y1 - 2022

N2 - Knowledge base (KB) embeddings have been shown to contain gender biases. In this paper, we study two questions regarding these biases: how to quantify them, and how to trace their origins in KB? Specifically, first, we develop two novel bias measures respectively for a group of person entities and an individual person entity. Evidence of their validity is observed by comparison with real-world census data. Second, we use the influence function to inspect the contribution of each triple in KB to the overall group bias. To exemplify the potential applications of our study, we also present two strategies (by adding and removing KB triples) to mitigate gender biases in KB embeddings.

AB - Knowledge base (KB) embeddings have been shown to contain gender biases. In this paper, we study two questions regarding these biases: how to quantify them, and how to trace their origins in KB? Specifically, first, we develop two novel bias measures respectively for a group of person entities and an individual person entity. Evidence of their validity is observed by comparison with real-world census data. Second, we use the influence function to inspect the contribution of each triple in KB to the overall group bias. To exemplify the potential applications of our study, we also present two strategies (by adding and removing KB triples) to mitigate gender biases in KB embeddings.

U2 - 10.18653/v1/2022.acl-long.98

DO - 10.18653/v1/2022.acl-long.98

M3 - Conference contribution

SP - 1381

EP - 1395

BT - Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)

A2 - Muresan, S.

A2 - Nakov, P.

A2 - Villavicencio, A.

PB - Association for Computational Linguistics

T2 - Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)

Y2 - 22 May 2022 through 27 May 2022

ER -

Understanding Gender Biases in Knowledge Base Embeddings

Abstract

Conference

Access to Document

Fingerprint

Cite this