Bias in word embeddings

Papakyriakopoulos, Orestis; Hegelich, Simon; Serrano, Juan Carlos Medina; Marco, Fabienne

doi:10.1145/3351095.3372843

Cited by 72 publications

(40 citation statements)

References 67 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…It has been proven that classifiers trained on biased word embeddings replicate the bias encoded in the word embeddings (Papakyriakopoulos et al, 2020 ). The authors trained a sentiment classifier on word embeddings that had before been proven to be biased.…”

Section: Discussionmentioning

confidence: 99%

“…Other approaches consider a mitigation at the level of the application that is using the word embeddings. For example, it has been shown that the mitigation at the level of a classifier that was trained on biased word embeddings was efficient (Papakyriakopoulos et al, 2020 ). The authors claim that the classifier learns further associations between the vectors, which are not considered when debiasing at the level of the word embeddings.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

A World Full of Stereotypes? Further Investigation on Origin and Gender Bias in Multi-Lingual Word Embeddings

Kurpicz-Briki

Leoni

2021

Front. Big Data

View full text Add to dashboard Cite

Publicly available off-the-shelf word embeddings that are often used in productive applications for natural language processing have been proven to be biased. We have previously shown that this bias can come in different forms, depending on the language and the cultural context. In this work, we extend our previous work and further investigate how bias varies in different languages. We examine Italian and Swedish word embeddings for gender and origin bias, and demonstrate how an origin bias concerning local migration groups in Switzerland is included in German word embeddings. We propose BiasWords, a method to automatically detect new forms of bias. Finally, we discuss how cultural and language aspects are relevant to the impact of bias on the application and to potential mitigation measures.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

A World Full of Stereotypes? Further Investigation on Origin and Gender Bias in Multi-Lingual Word Embeddings

Kurpicz-Briki

Leoni

2021

Front. Big Data

View full text Add to dashboard Cite

show abstract

“…On the other hand, Merenda et al (2018) have shown the effectiveness of using messages from potentially abusive-oriented on-line communities to generate so-called hate embeddings. More recently, Papakyriakopoulos et al (2020) have shown that biased word embeddings can be beneficial. We follow the idea of exploiting biased embeddings by creating them using messages from banned communities in Reddit.…”

Section: Hatebert: Re-training Bert With Abusive Online Communitiesmentioning

confidence: 99%

HateBERT: Retraining BERT for Abusive Language Detection in English

Caselli¹,

Basile²,

Mitrović³

et al. 2021

Proceedings of the 5th Workshop on Online Abuse and Harms (WOAH 2021)

152

View full text Add to dashboard Cite

We introduce HateBERT, a re-trained BERT model for abusive language detection in English. The model was trained on RAL-E, a large-scale dataset of Reddit comments in English from communities banned for being offensive, abusive, or hateful that we have curated and made available to the public. We present the results of a detailed comparison between a general pre-trained language model and the retrained version on three English datasets for offensive, abusive language and hate speech detection tasks. In all datasets, HateBERT outperforms the corresponding general BERT model. We also discuss a battery of experiments comparing the portability of the fine-tuned models across the datasets, suggesting that portability is affected by compatibility of the annotated phenomena.

show abstract

“…Similarly to how long existing stereotypes are deep-rooted in word embeddings (Papakyriakopoulos et al, 2020;Garg et al, 2018), PTLMs have also been shown to recreate stereotypical content due to the nature of their training data (Sheng et al, 2019) Different probing experiments have been proposed to study the drawbacks of PTLMs in areas such as the biomedical domain (Jin et al, 2019), syntax (Hewitt and Manning, 2019;Marvin and Linzen, 2018), semantic and syntactic sentence structures (Tenney et al, 2019), prenomial anaphora (Sorodoc et al, 2020), common-sense (Petroni et al, 2019), gender bias (Kurita et al, 2019), and typicality in judgement (Misra et al, 2021). Except for Hutchinson et al (2020) who examine what words BERT generate in some fill-in-the-blank experiments with regard to people with disabilities, and more recently Nozza et al (2019) who assess hurtful auto-completion by multilingual PTLMs, we are not aware of other strategies designed to estimate toxic content in PTLMs with regard to several social groups.…”

Section: Related Workmentioning

confidence: 99%

“…Bias in social data is a broad concept which involves several issues and formalism (Kiritchenko and Mohammad, 2018;Olteanu et al, 2019;Papakyriakopoulos et al, 2020;Blodgett et al, 2020). For instance, Shah et al (2020) present a framework to predict the origin of different types of bias including label bias (Sap et al, 2019a), selection bias (Garimella et al, 2019;Ousidhoum et al, 2020), model overamplification (Zhao et al, 2017), and semantic bias (Garg et al, 2018).…”

Section: Related Workmentioning

confidence: 99%

Probing Toxic Content in Large Pre-Trained Language Models

Ousidhoum¹,

Zhao²,

Fang³

et al. 2021

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Confer

View full text Add to dashboard Cite

Large pre-trained language models (PTLMs) have been shown to carry biases towards different social groups which leads to the reproduction of stereotypical and toxic content by major NLP systems. We propose a method based on logistic regression classifiers to probe English, French, and Arabic PTLMs and quantify the potentially harmful content that they convey with respect to a set of templates. The templates are prompted by a name of a social group followed by a cause-effect relation. We use PTLMs to predict masked tokens at the end of a sentence in order to examine how likely they enable toxicity towards specific communities. We shed the light on how such negative content can be triggered within unrelated and benign contexts based on evidence from a large-scale study, then we explain how to take advantage of our methodology to assess and mitigate the toxicity transmitted by PTLMs.

show abstract

Bias in word embeddings

Cited by 72 publications

References 67 publications

A World Full of Stereotypes? Further Investigation on Origin and Gender Bias in Multi-Lingual Word Embeddings

A World Full of Stereotypes? Further Investigation on Origin and Gender Bias in Multi-Lingual Word Embeddings

HateBERT: Retraining BERT for Abusive Language Detection in English

Probing Toxic Content in Large Pre-Trained Language Models

Contact Info

Product

Resources

About