Leveraging Bias in Pre-Trained Word Embeddings for Unsupervised Microaggression Detection

Tolúlope' Ògúnremí; Nazanin Sabri; Valerio Basile; Tommaso Caselli

Leveraging Bias in Pre-Trained Word Embeddings for Unsupervised Microaggression Detection

Tolúlope' Ògúnremí, Nazanin Sabri, Valerio Basile, Tommaso Caselli

Computational Linguistics (CL)

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

76 Downloads (Pure)

Abstract

Microaggressions are subtle manifestations of bias (Breitfeller et al., 2019).
These demonstrations of bias can often be classified as a subset of abusive language. However, not as much focus has been placed on the recognition of these instances. As a result, limited data is available on the topic, and only in English. Being able to detect microaggressions without the need for labeled data would be advantageous since it would allow content moderation also for languages lacking annotated data. In this study, we introduce an unsupervised method to detect microaggressions in natural language expressions.
The algorithm relies on pre-trained word-embeddings, leveraging the bias encoded in the model in order to detect microaggressions in unseen textual instances. We test the method on a dataset of racial and gender-based microaggressions, reporting promising results. We further run the algorithm on out-of-domain unseen data with the purpose of bootstrapping corpora of
microaggressions “in the wild”, and discuss the benefits and drawbacks of our
proposed method.

Original language	English
Title of host publication	Proceedings of the Eighth Italian Conference on Computational Linguistics
Editors	Elisabetta Fersini, Marco Passarotti, Viviana Patti
Publisher	CEUR Workshop Proceedings (CEUR-WS.org)
Number of pages	7
Publication status	Published - 2021
Event	Italian Conference on Computational Linguistics 2021: CLiC-it 2021 - Milan, Italy Duration: 26-Jan-2022 → 28-Jan-2022 Conference number: 8

Conference

Conference	Italian Conference on Computational Linguistics 2021
Country/Territory	Italy
City	Milan
Period	26/01/2022 → 28/01/2022

Keywords

micro-aggression
hate speech
NLP

Access to Document

Leveraging Bias in Pre-Trained Word Embeddings for Unsupervised Microaggression DetectionFinal publisher's version, 394 KBLicence: CC BY

Handle.net

Persistent link

Best Student Paper Award
Ògúnremí, T. (Recipient), Sabri, N. (Recipient), Basile, V. (Recipient) & Caselli, T. (Recipient), 2022
Prize: National/international honour › Academic

Cite this

@inproceedings{0706cd1e31bd4a289ba9f7079e457d06,

title = "Leveraging Bias in Pre-Trained Word Embeddings for Unsupervised Microaggression Detection",

abstract = "Microaggressions are subtle manifestations of bias (Breitfeller et al., 2019).These demonstrations of bias can often be classified as a subset of abusive language. However, not as much focus has been placed on the recognition of these instances. As a result, limited data is available on the topic, and only in English. Being able to detect microaggressions without the need for labeled data would be advantageous since it would allow content moderation also for languages lacking annotated data. In this study, we introduce an unsupervised method to detect microaggressions in natural language expressions.The algorithm relies on pre-trained word-embeddings, leveraging the bias encoded in the model in order to detect microaggressions in unseen textual instances. We test the method on a dataset of racial and gender-based microaggressions, reporting promising results. We further run the algorithm on out-of-domain unseen data with the purpose of bootstrapping corpora ofmicroaggressions “in the wild”, and discuss the benefits and drawbacks of ourproposed method.",

keywords = "micro-aggression, hate speech, NLP",

author = "Tol{\'u}lope' {\`O}g{\'u}nrem{\'i} and Nazanin Sabri and Valerio Basile and Tommaso Caselli",

year = "2021",

language = "English",

editor = "Elisabetta Fersini and Marco Passarotti and Viviana Patti",

booktitle = "Proceedings of the Eighth Italian Conference on Computational Linguistics",

publisher = "CEUR Workshop Proceedings (CEUR-WS.org)",

note = "Italian Conference on Computational Linguistics 2021 : CLiC-it 2021 ; Conference date: 26-01-2022 Through 28-01-2022",

}

Ògúnremí, T, Sabri, N, Basile, V & Caselli, T 2021, Leveraging Bias in Pre-Trained Word Embeddings for Unsupervised Microaggression Detection. in E Fersini, M Passarotti & V Patti (eds), Proceedings of the Eighth Italian Conference on Computational Linguistics. CEUR Workshop Proceedings (CEUR-WS.org), Italian Conference on Computational Linguistics 2021, Milan, Italy, 26/01/2022.

Leveraging Bias in Pre-Trained Word Embeddings for Unsupervised Microaggression Detection. / Ògúnremí, Tolúlope'; Sabri, Nazanin; Basile, Valerio et al.
Proceedings of the Eighth Italian Conference on Computational Linguistics. ed. / Elisabetta Fersini; Marco Passarotti; Viviana Patti. CEUR Workshop Proceedings (CEUR-WS.org), 2021.

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

TY - GEN

T1 - Leveraging Bias in Pre-Trained Word Embeddings for Unsupervised Microaggression Detection

AU - Ògúnremí, Tolúlope'

AU - Sabri, Nazanin

AU - Basile, Valerio

AU - Caselli, Tommaso

N1 - Conference code: 8

PY - 2021

Y1 - 2021

N2 - Microaggressions are subtle manifestations of bias (Breitfeller et al., 2019).These demonstrations of bias can often be classified as a subset of abusive language. However, not as much focus has been placed on the recognition of these instances. As a result, limited data is available on the topic, and only in English. Being able to detect microaggressions without the need for labeled data would be advantageous since it would allow content moderation also for languages lacking annotated data. In this study, we introduce an unsupervised method to detect microaggressions in natural language expressions.The algorithm relies on pre-trained word-embeddings, leveraging the bias encoded in the model in order to detect microaggressions in unseen textual instances. We test the method on a dataset of racial and gender-based microaggressions, reporting promising results. We further run the algorithm on out-of-domain unseen data with the purpose of bootstrapping corpora ofmicroaggressions “in the wild”, and discuss the benefits and drawbacks of ourproposed method.

AB - Microaggressions are subtle manifestations of bias (Breitfeller et al., 2019).These demonstrations of bias can often be classified as a subset of abusive language. However, not as much focus has been placed on the recognition of these instances. As a result, limited data is available on the topic, and only in English. Being able to detect microaggressions without the need for labeled data would be advantageous since it would allow content moderation also for languages lacking annotated data. In this study, we introduce an unsupervised method to detect microaggressions in natural language expressions.The algorithm relies on pre-trained word-embeddings, leveraging the bias encoded in the model in order to detect microaggressions in unseen textual instances. We test the method on a dataset of racial and gender-based microaggressions, reporting promising results. We further run the algorithm on out-of-domain unseen data with the purpose of bootstrapping corpora ofmicroaggressions “in the wild”, and discuss the benefits and drawbacks of ourproposed method.

KW - micro-aggression

KW - hate speech

KW - NLP

M3 - Conference contribution

BT - Proceedings of the Eighth Italian Conference on Computational Linguistics

A2 - Fersini, Elisabetta

A2 - Passarotti, Marco

A2 - Patti, Viviana

PB - CEUR Workshop Proceedings (CEUR-WS.org)

T2 - Italian Conference on Computational Linguistics 2021

Y2 - 26 January 2022 through 28 January 2022

ER -

Leveraging Bias in Pre-Trained Word Embeddings for Unsupervised Microaggression Detection

Abstract

Conference

Keywords

Access to Document

Handle.net

Fingerprint

Prizes

Best Student Paper Award

Cite this