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
Queens are Powerful too: Mitigating Gender Bias in Dialogue Generation
[article]
2020
arXiv
pre-print
Preserved Fulltext
Other Versions
We show that our proposed techniques mitigate gender bias in LIGHT by balancing the genderedness of generated dialogue utterances and are particularly effective in combination. ...
We analyze gender bias in dialogue data, and examine how this bias is actually amplified in subsequent generative chit-chat dialogue models. ...
a general framework for mitigating bias in generative dialogue. ...
arXiv:1911.03842v2
fatcat:resliw5syrfgzavwddk3ixp4nq
Societal Biases in Language Generation: Progress and Challenges
[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.
Other Versions
To better understand these challenges, we present a survey on societal biases in language generation, focusing on how data and techniques contribute to biases and progress towards reducing biases. ...
Language generation presents unique challenges for biases in terms of direct user interaction and the structure of decoding techniques. ...
Queens are powerful too: Mitigating gender bias in
dialogue generation. ...
arXiv:2105.04054v3
fatcat:dhwma4hvfbf7jke3lt227qb73i
RedditBias: A Real-World Resource for Bias Evaluation and Debiasing of Conversational 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.
Recent work has predominantly focused on measuring and mitigating bias in pretrained language models. ...
In this work, we present RedditBias, the first conversational data set grounded in the actual human conversations from Reddit, allowing for bias measurement and mitigation across four important bias dimensions ...
Queens are powerful too: Mitigating gender bias in
dialogue generation. ...
arXiv:2106.03521v1
fatcat:s43dffsi25goppqnb742qxcf7e
Language (Technology) is Power: A Critical Survey of "Bias" in NLP
[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.
Other Versions
We survey 146 papers analyzing "bias" in NLP systems, finding that their motivations are often vague, inconsistent, and lacking in normative reasoning, despite the fact that analyzing "bias" is an inherently ...
We further find that these papers' proposed quantitative techniques for measuring or mitigating "bias" are poorly matched to their motivations and do not engage with the relevant literature outside of ...
We excluded from the "Vague/unstated" category of motivations the papers that participated in the Gendered Ambiguous Pronoun (GAP) Shared Task at the First ACL Workshop on Gender Bias in NLP. ...
arXiv:2005.14050v2
fatcat:nyzr4fj5gne55jzhzm2fy5xeqa
The Arabic Parallel Gender Corpus 2.0: Extensions and Analyses
[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.
Much of the research on this issue has focused on mitigating gender bias in English NLP models and systems. ...
Gender bias in natural language processing (NLP) applications, particularly machine translation, has been receiving increasing attention. ...
We acknowledge that by limiting the choice of gender expression to the grammatical gender choices in Arabic, we exclude other alternatives such as non-binary gender or no-gender expressions. ...
arXiv:2110.09216v1
fatcat:twukzdxbbbgvrnicwffrpf24ky
Multi-Dimensional Gender Bias Classification
[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.
Machine learning models are trained to find patterns in data. NLP models can inadvertently learn socially undesirable patterns when training on gender biased text. ...
In this work, we propose a general framework that decomposes gender bias in text along several pragmatic and semantic dimensions: bias from the gender of the person being spoken about, bias from the gender ...
Queens are powerful too: Mitigating gender bias in
dialogue generation.
Emily Dinan, Samuel Humeau, Bharath Chintagunta,
and Jason Weston. 2019b. ...
arXiv:2005.00614v1
fatcat:o3lgzjeouvhepmp6bkmw2jk7jm
Mitigating Language-Dependent Ethnic Bias in BERT
[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.
Other Versions
In this paper, we study ethnic bias and how it varies across languages by analyzing and mitigating ethnic bias in monolingual BERT for English, German, Spanish, Korean, Turkish, and Chinese. ...
BERT and other large-scale language models (LMs) contain gender and racial bias. ...
Queens are powerful too: Mitigating gender bias in
dialogue generation. ...
arXiv:2109.05704v2
fatcat:yfzqtcpxwbcgzcxd4xbfef53fi
Just Say No: Analyzing the Stance of Neural Dialogue Generation in Offensive Contexts
[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.
Other Versions
Finally, we quantify the effectiveness of controllable text generation (CTG) methods to mitigate the tendency of neural dialogue models to agree with offensive comments. ...
Dialogue models trained on human conversations inadvertently learn to generate toxic responses. ...
This material is based in part on research sponsored by the NSF (IIS-1845670) and DARPA via the ARO (W911NF-17-C-0095). ...
arXiv:2108.11830v2
fatcat:vrrhe53npndghmw2sla4i7ah4i
Adopted vs. Adapted EFL Textbooks in Iranian Context: Focusing on Gender Representation and Teacher Beliefs
2013
Journal of Language Teaching and Research
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 2017; you can also visit the original URL.
The file type is application/pdf.
Studies show that most of textbooks are biased in a way that they present stereotypical female images. ...
Implications of the study are discussed. ...
They found that there was some gender bias in the textbook dialogues but on the whole, the bias was not extreme. ...
doi:10.4304/jltr.4.4.791-800
fatcat:qro3melk6fgibbd23yrcs5frqe
The Trans Panic Defense: Heteronormativity, and the Murder of Transgender Women
2014
Social Science Research Network
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 2016; you can also visit the original URL.
The file type is application/pdf.
The law, however, generally disapproves of acts of violence that are motivated by a desire to enforce gender norms. 16. ...
"This is just too much." ...
doi:10.2139/ssrn.2430390
fatcat:ex4twcyaunc4tlr5inciupyj6y
READING SIR THOMAS SMITH'S DE REPUBLICA ANGLORUM AS PROTESTANT APOLOGETIC
1999
The Historical Journal
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 2004; you can also visit the original URL.
The file type is application/pdf.
The first establishes Smith's ideological concerns and convictions in Edward VI's reign and in the early years of Elizabeth's. ...
The conclusion locates Smith's text in relation to one further context : Claude de Seyssel's The monarchy of France and its use by French Huguenot theorists in the s. ...
Protestant national identity is too often overlooked.$" But Smith actually wrote a tract urging the queen to marry in , the Dialogue concerning the queen's marriage.$# Like De republica anglorum, ...
doi:10.1017/s0018246x99008730
fatcat:ogi5nddbsrhdvnr4uedqpvfhiu
Just Cognition: Scientific Research on Bias and Some Implications for Legal Procedure and Decision‐Making
2019
Modern law review
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.
Judges have insisted that some potential sources of bias are not open to review and maintained the idea that they are peculiarly resistant to bias through legal training and judicial experience. ...
Bias is believed to threaten the administration of justice and the legitimacy of legal decision-making, particularly public confidence in the courts. ...
done so.' 89 Those who believe themselves to be less biased than others generally are not. ...
doi:10.1111/1468-2230.12424
fatcat:yirkvttiffcbjivi6dm4b6vwcu
Towards Neural Programming Interfaces
[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.
Other Versions
It is notoriously difficult to control the behavior of artificial neural networks such as generative neural language models. ...
Importantly, no permanent changes are made to the weights of the original model, allowing us to re-purpose pretrained models for new tasks without overwriting any aspect of the language model. ...
Acknowledgments and Disclosure of Funding We would like to acknowledge the significant contribution of our collaborators Junseong Ahn and Benjamin Murdoch, whose discussion and insights were invaluable in ...
arXiv:2012.05983v2
fatcat:4tpggy33uzdu3myvq2twly3quy
Beyond the martial façade: gender, heritage and medieval castles
2019
International Journal of Heritage Studies (IJHS)
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.
The enduring narrative of militarism at medieval castles sites is discussed, together with issues of authenticity in relation to the historical record, which is in itself biased and inherently gendered ...
Gendered interpretations are rare both within castle-studies and heritage discourses on medieval castles. ...
It illustrated too, the biases that are present within our workin how we are telling the story to the public today. ...
doi:10.1080/13527258.2019.1636119
fatcat:pag6ewyycbfpplvjdbph3bwqyu
In Defense of Safe Spaces: A Phenomenological Account
2018
Atlantis: A Womens Studies Journal
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.
through power discrepancies are erased. ...
Moreover, the anxieties that arise in the expectation of intrusion of the unsafe reflect gendered, sexual, and racialized power relations that are effects of a system of structural violence (Koskela 1997 ...
I include this example as it also shows that white male students are not silenced in safe(r) space classrooms. ...
doi:10.7202/1064069ar
fatcat:rbuqnzkn7vgw3jnijepxhz6jzm
« Previous
Showing results 1 — 15 out of 2,078 results