A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Filters
On Measuring Social Biases in Sentence Encoders
2019
Proceedings of the 2019 Conference of the North
Preserved Fulltext
We conclude by proposing directions for future work on measuring bias in sentence encoders. ...
Accordingly, we extend the Word Embedding Association Test to measure bias in sentence encoders. ...
The views and conclusions contained in this publication are those of the authors and should not be interpreted as representing official policies or endorsements of IARPA, DARPA, NSF, or the U.S. ...
doi:10.18653/v1/n19-1063
dblp:conf/naacl/MayWBBR19
fatcat:l4qv7oeogzefxfq6ni3geqsez4
On Measuring Social Biases in Sentence Encoders
[article]
2019
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2019; you can also visit the original URL.
The file type is application/pdf.
We conclude by proposing directions for future work on measuring bias in sentence encoders. ...
Accordingly, we extend the Word Embedding Association Test to measure bias in sentence encoders. ...
The views and conclusions contained in this publication are those of the authors and should not be interpreted as representing official policies or endorsements of IARPA, DARPA, NSF, or the U.S. ...
arXiv:1903.10561v1
fatcat:i3hy7fhus5d3liidcicmbxl37q
Towards Debiasing Sentence Representations
[article]
2020
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
In this paper, we investigate the presence of social biases in sentence-level representations and propose a new method, Sent-Debias, to reduce these biases. ...
play in shaping social biases and stereotypes. ...
RS was supported in part by US Army, ONR, Apple, and NSF IIS1763562. ...
arXiv:2007.08100v1
fatcat:szeqyyiikzcmhdihuea2wowd54
Unmasking the Mask – Evaluating Social Biases in Masked Language Models
2022
PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
) to evaluate tokens based on their importance in a sentence. ...
overestimate the measured biases and are heavily influenced by the unmasked tokens in the context. ...
pair is socially biased using an MLM-based bias scoring method. 9 We split sentence pairs in the CP dataset into two groups depending on whether a sentence pair has received more than three biased ratings ...
doi:10.1609/aaai.v36i11.21453
fatcat:7fjejpt42jgblbdub27dackshi
Sense Embeddings are also Biased–Evaluating Social Biases in Static and Contextualised Sense Embeddings
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Other Versions
One sense of an ambiguous word might be socially biased while its other senses remain unbiased. ...
We create a benchmark dataset for evaluating the social biases in sense embeddings and propose novel sense-specific bias evaluation measures. ...
Using SSSB, we show that the proposed bias evaluation measures for sense embeddings capture different types of social biases encoded in existing SoTA sense embeddings. ...
arXiv:2203.07523v2
fatcat:ryvpr24bsnfizgyofqqvpe3ehm
Assessing Social and Intersectional Biases in Contextualized Word Representations
[article]
2019
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
These word representations are conditioned on their context within a sentence, and can also be used to encode the entire sentence. ...
In this paper, we analyze the extent to which state-of-the-art models for contextual word representations, such as BERT and GPT-2, encode biases with respect to gender, race, and intersectional identities ...
The first author would also like to thank John Lafferty for inspiring interest in bias in word representations. ...
arXiv:1911.01485v1
fatcat:j7k3ubdnsnbytk5ezqp3yprqi4
Mapping the Multilingual Margins: Intersectional Biases of Sentiment Analysis Systems in English, Spanish, and Arabic
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
We use these tools to measure gender, racial, ethnic, and intersectional social biases across five models trained on emotion regression tasks in English, Spanish, and Arabic. ...
social biases in natural language processing. ...
Our decision to not fine-tune does decrease performance on downstream tasks but is prudent given the risk of overfitting on a small training set and our interest in studying the social biases encoded in ...
arXiv:2204.03558v1
fatcat:vmqtnfefibfwblzr4nnuviv5pe
Socially Aware Bias Measurements for Hindi Language Representations
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Other Versions
In this work, we investigate biases present in Hindi language representations with focuses on caste and religion-associated biases. ...
These biases are studied extensively, but with a primary focus on English language representations and biases common in the context of Western society. ...
May et al. (2019) propose SEAT (Sentence Embedding Association Test) for measuring bias in sentence encoders. ...
arXiv:2110.07871v2
fatcat:q7qv4hvfl5cl7mcmvz3iksynze
Measuring Social Biases in Grounded Vision and Language Embeddings
[article]
2020
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
This is despite the fact that vision and language can suffer from different biases, which one might hope could attenuate the biases in both. ...
We generalize the notion of social biases from language embeddings to grounded vision and language embeddings. ...
., the encoding of a word in context in the sentence. ...
arXiv:2002.08911v1
fatcat:b4dqgteahjfgpafg4vvzpwxtt4
VAST: The Valence-Assessing Semantics Test for Contextualizing Language Models
2022
PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
that non-semantic structures in LMs also mask social biases. ...
We also show that multiply tokenized words are not semantically encoded until layer 8, where they achieve Pearson's rho of .46, indicating the presence of an encoding process for multiply tokenized words ...
After isolating semantics, VAST measures social biases, an important step for assessing the potential for harmful associations to manifest in downstream tasks. ...
doi:10.1609/aaai.v36i10.21400
fatcat:247uzckqezaohmu4lbyzbajxou
A Survey on Bias and Fairness in Natural Language Processing
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
In this survey, we analyze the origins of biases, the definitions of fairness, and how different subfields of NLP mitigate bias. ...
cycle in many settings. ...
Different variations of A, B, X, Y are used in different papers.
Sentence Encoder Association Test This is a simple generalization of the WEAT test for sentence encoders. ...
arXiv:2204.09591v1
fatcat:7arytn5jsrcdtatcnimnpzdmy4
Social relevance enhances memory for impressions in older adults
2012
Memory
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2015; you can also visit the original URL.
The file type is application/pdf.
This may suggest that while age-related biases might occur in an implicit encoding context, making judgements in an explicit task may reduce biases towards positive information. ...
(Folstein, Folstein, & McHugh, 1975 ) (M 028.75, SD 0 1.07), and were characterised on cognitive measures to ensure comparability to other samples in the literature. ...
doi:10.1080/09658211.2012.660956
pmid:22364168
pmcid:PMC3371644
fatcat:aingienn75emrffpqykiajnuai
Towards Understanding and Mitigating Social Biases in Language Models
[article]
2021
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
As machine learning methods are deployed in real-world settings such as healthcare, legal systems, and social science, it is crucial to recognize how they shape social biases and stereotypes in these sensitive ...
With these tools, we propose steps towards mitigating social biases during text generation. ...
RS is supported in part by NSF IIS1763562 and ONR Grant N000141812861. ...
arXiv:2106.13219v1
fatcat:yjkjuktjyjbejjp2axyc3wprhy
Addressing Age-Related Bias in Sentiment Analysis
2019
Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Recent studies have identified various forms of bias in language-based models, raising concerns about the risk of propagating social biases against certain groups based on sociodemographic factors (e.g ...
Our results show significant age bias is encoded in the outputs of many sentiment analysis algorithms and word embeddings, and we can alleviate this bias by manipulating training data. ...
Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence ...
doi:10.24963/ijcai.2019/852
dblp:conf/ijcai/DiazJLPG19
fatcat:lfntknjvpbfv7ovjmbeoye3ufy
A neural mechanism of first impressions
2009
Nature Neuroscience
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2011; you can also visit the original URL.
The file type is application/pdf.
Neuroimaging revealed that responses in the amygdala and the posterior cingulate cortex (PCC) were stronger while encoding social information that was consistent, relative to inconsistent, with subsequent ...
These findings provide evidence for encoding differences on the basis of subsequent evaluations, suggesting that the amygdala and PCC are important for forming first impressions. ...
To test for these memory biases, we used a recognition accuracy measure that was defined as correct recognition responses divided by the total number of sentences in category. ...
doi:10.1038/nn.2278
pmid:19270690
fatcat:7zu5orainzakncc3v5fge6rwri
« Previous
Showing results 1 — 15 out of 15,356 results