42 research outputs found
An image speaks a thousand words, but can everyone listen? On image transcreation for cultural relevance
No full textGiven the rise of multimedia content, human translators increasingly focus on culturally adapting not only words but also other modalities such as images to convey the same meaning. While several applications stand to benefit from this, machine translation systems remain confined to dealing with language in speech and text. In this work, we take a first step towards translating images to make them culturally relevant. First, we build three pipelines comprising state-of-the-art generative models to do the task. Next, we build a two-part evaluation dataset: i) concept: comprising 600 images that are cross-culturally coherent, focusing on a single concept per image, and ii) application: comprising 100 images curated from real-world applications. We conduct a multi-faceted human evaluation of translated images to assess for cultural relevance and meaning preservation. We find that as of today, image-editing models fail at this task, but can be improved by leveraging LLMs and retrievers in the loop. Best pipelines can only translate 5% of images for some countries in the easier concept dataset and no translation is successful for some countries in the application dataset, highlighting the challenging nature of the task. Our code and data is released here: https://github.com/simran-khanuja/image-transcreation
DeMuX: Data-efficient Multilingual Learning
No full textWe consider the task of optimally fine-tuning pre-trained multilingual
models, given small amounts of unlabelled target data and an annotation budget.
In this paper, we introduce DEMUX, a framework that prescribes the exact
data-points to label from vast amounts of unlabelled multilingual data, having
unknown degrees of overlap with the target set. Unlike most prior works, our
end-to-end framework is language-agnostic, accounts for model representations,
and supports multilingual target configurations. Our active learning strategies
rely upon distance and uncertainty measures to select task-specific neighbors
that are most informative to label, given a model. DeMuX outperforms strong
baselines in 84% of the test cases, in the zero-shot setting of disjoint source
and target language sets (including multilingual target pools), across three
models and four tasks. Notably, in low-budget settings (5-100 examples), we
observe gains of up to 8-11 F1 points for token-level tasks, and 2-5 F1 for
complex tasks. Our code is released here:
https://github.com/simran-khanuja/demux
The Light We Give: Sikh Wisdom for Cultivating Empathy and Justice
No full textGrowing up in South Texas, Dr. Simran Jeet Singh and his brothers confronted racism daily. As a turbaned, bearded, brown-skinned Sikh, he continued to face prejudice and hate in college and beyond. Simran chose to be defined not by the negativity that often surrounded him but by the Sikh teachings of love and justice that he grew up with. Delving deep into these core tenets of Sikh wisdom, he has sought to embrace an outlook that guides us to see the good in everyone and to forge a path of positivity, connection, and service—a way of life that so many of us are seeking in today’s world.
We all say that we choose love over hate. But when tested, we realize that it’s easier said than done and that our empathy for others is not rooted deeply enough. As a turbaned and bearded Sikh man, Simran has been subjected to racism his whole life. He has been working on the frontlines of hate violence for more than a decade. And yet, he has managed to avoid falling into the toxic trap of hate and anger. In this lecture, drawing on his recent book The Light We Give, he will draw from his personal experiences and from hate incidents he has witnessed firsthand to share the wisdom he has gained on what it really takes to choose love over hate.
Simran Jeet Singh, Ph.D., is the Executive Director of the Religion & Society Program at the Aspen Institute and the author of the national bestseller The Light We Give: How Sikh Wisdom Can Transform Your Life (Riverhead, Penguin Random House). Simran\u27s thought leadership on bias, empathy, and justice extends across corporate, university, and government settings. He is an Atlantic Fellow for Racial Equity with Columbia University and the Nelson Mandela Foundation, a Soros Equality Fellow with the Open Society Foundations, a Visiting Lecturer at Union Seminary, and a Senior Advisor on Equity and Inclusion for YSC Consulting, part of Accenture.
Organized and hosted by the Interfaith Fellows Program of the Jay Phillips Center for Interreligious Studies at the University of St. Thomas and the Minnesota Multifaith Network in collaboration with the Lutheran Center for Faith, Values, and Community at St. Olaf College and the Interfaith Institute at Augsburg University. Cosponsored by Minnesota Multifaith Network, and the Office of Diversity, Equity and Inclusion, the College of Arts and Sciences, the Diversity Activities Board (DAB), and the Department of Theology at the University of St. Thomas. Funded, in part, by generous grants from the Arthur Vining Davis Foundations, the Jay and Rose Phillips Family Foundation of Minnesota, and the Center for Faculty Development at the University of St. Thomas
Keynote Address: The Light We Give: Sikh Wisdom for Cultivating Empathy and Justice
No full textGrowing up in South Texas, Dr. Simran Jeet Singh and his brothers confronted racism daily. As a turbaned, bearded, brown-skinned Sikh, he continued to face prejudice and hate in college and beyond. Simran chose to be defined not by the negativity that often surrounded him but by the Sikh teachings of love and justice that he grew up with. Delving deep into these core tenets of Sikh wisdom, he has sought to embrace an outlook that guides us to see the good in everyone and to forge a path of positivity, connection, and service—a way of life that so many of us are seeking in today’s world.
We all say that we choose love over hate. But when tested, we realize that it’s easier said than done and that our empathy for others is not rooted deeply enough. As a turbaned and bearded Sikh man, Simran has been subjected to racism his whole life. He has been working on the frontlines of hate violence for more than a decade. And yet, he has managed to avoid falling into the toxic trap of hate and anger. In this lecture, drawing on his recent book The Light We Give, he will draw from his personal experiences and from hate incidents he has witnessed firsthand to share the wisdom he has gained on what it really takes to choose love over hate.
Simran Jeet Singh, Ph.D., is the Executive Director of the Religion & Society Program at the Aspen Institute and the author of the national bestseller The Light We Give: How Sikh Wisdom Can Transform Your Life (Riverhead, Penguin Random House). Simran\u27s thought leadership on bias, empathy, and justice extends across corporate, university, and government settings. He is an Atlantic Fellow for Racial Equity with Columbia University and the Nelson Mandela Foundation, a Soros Equality Fellow with the Open Society Foundations, a Visiting Lecturer at Union Seminary, and a Senior Advisor on Equity and Inclusion for YSC Consulting, part of Accenture.
Organized and hosted by the Interfaith Fellows Program of the Jay Phillips Center for Interreligious Studies at the University of St. Thomas and the Minnesota Multifaith Network in collaboration with the Lutheran Center for Faith, Values, and Community at St. Olaf College and the Interfaith Institute at Augsburg University. Cosponsored by Minnesota Multifaith Network, and the Office of Diversity, Equity and Inclusion, the College of Arts and Sciences, and the Department of Theology at the University of St. Thomas. the , and in collaboration with the Office of Diversity, Equity and Inclusion at the University of St. Thomas. Funded, in part, by generous grants from the Arthur Vining Davis Foundations, the Jay and Rose Phillips Family Foundation of Minnesota, and the Center for Faculty Development at the University of St. Thomas
Critical Anaylsis on the Effects of Triple Talaq, the Plight of Women, its Impact on the Society Muslim Community
No full textToday, the issues of women rights in muslim personal law is highly controversial. Specially, muslim women rights relating to triple talaq, inheritance, maintenance has got much attention nowadays. A muslim man can divorce his wife by prouncing three times talaq. When husband clearly mentions it is called as express talaq. After that husband and wife cannot be together back until wife marries someone else. The legal decisions are based on the norms mentioned in quaran therefore, certain anomalies need to be eradicated by giving true essence of holy quaran for the benefit of muslim women's right. There is three types of talaq namely, unlike other religion marriage is viewed as sacrament but, under, muslim law it is civil and social contract. Talaq ul sunnat sanctioned by prophet is sub divided into Talaq e ehsan, Talaq hasan, Talaq e biddat. The current debate on triple talaq, centred on the Sharaya Bano and several other petitions which considers no aspect of Islamic personal laws which amounts to violate the spirit of constitution. The whole triple talaq has become a battleground for the culture vs social debate. In this paper the author deals with the question of triple talaq in the light of the recent petition filed in the Supreme Court for declaring such talaq invalid. The author argues that there is an already existing legal precedent established by the apex court with respect triple talaq which should be followed instead of resorting in aggressive approach which may become dominant to muslim women themselves. This research paper analyze to attempt the on going implications on triple talaq, muslim personal law and solutions to empower muslim women. Simran Chhallani "Critical Anaylsis on the Effects of Triple Talaq, the Plight of Women, its Impact on the Society Muslim Community" Published in International Journal of Trend in Scientific Research and Development (ijtsrd), ISSN: 2456-6470, Volume-2 | Issue-5 , August 2018, URL: https://www.ijtsrd.com/papers/ijtsrd16996.pd
Nonprofit Youth Engagement: A Normalized Industrial-Complex
No full textThe 'Disrupting the Talent Pipeline: Youth Engagement & The Nonprofit-Industrial Complex' curriculum is intended to serve as an accessible resource to community that can be practically applied. This resource seeks to highlight key aspects of the experiences of 'multiply-marginalized' youth/young people navigating the nonprofit-sector, and its relationship to the industrial-complex that it perpetuates. The curriculum is informed by lived experiences as well as existing literature - offering both core content and activities/additional resources to guide its application in various community contexts.Not peer reviewe
Evaluating the Diversity, Equity and Inclusion of NLP Technology: A Case Study for Indian Languages
No full textIn order for NLP technology to be widely applicable, fair, and useful, it
needs to serve a diverse set of speakers across the world's languages, be
equitable, i.e., not unduly biased towards any particular language, and be
inclusive of all users, particularly in low-resource settings where compute
constraints are common. In this paper, we propose an evaluation paradigm that
assesses NLP technologies across all three dimensions. While diversity and
inclusion have received attention in recent literature, equity is currently
unexplored. We propose to address this gap using the Gini coefficient, a
well-established metric used for estimating societal wealth inequality. Using
our paradigm, we highlight the distressed state of current technologies for
Indian (IN) languages (a linguistically large and diverse set, with a varied
speaker population), across all three dimensions. To improve upon these
metrics, we demonstrate the importance of region-specific choices in model
building and dataset creation, and more importantly, propose a novel,
generalisable approach to optimal resource allocation during fine-tuning.
Finally, we discuss steps to mitigate these biases and encourage the community
to employ multi-faceted evaluation when building linguistically diverse and
equitable technologies.Comment: Accepted to EACL Findings, 202
Dependency Parser for Bengali-English Code-Mixed Data enhanced with a Synthetic Treebank
No full textRace through the finish line with your customers: Customer Segmentation and Profiling of CredRev
No full textDuring a semester long research project, our research team analyzed the CRM of CredRev, an auto financing company based in Kelowna, BC with customers all over BC. Our research objectives were to perform customer segmentation and profiling for the chosen organization, promote greater CRM strategies with the use of customer databases, and increase the amount of customers for CredRev by further understanding their customer segments and profiles.This poster won the Vice-President, Students award (2020). Supervisor: Dr. David Dobson, School of Business
Data Model for Computer Vision Explainability, Fairness, and Robustness
No full textIn recent years, there has been a growing interest among researchers in the explainability, fairness, and robustness of Computer Vision models. While studies have explored the usability of these models for end users, limited research has delved into the challenges and requirements faced by researchers investigating these requirements. This study addresses this gap through a mixed-method approach, involving 20 semi-structured interviews with researchers and a comprehensive literature analysis.Through this investigation, we have identified a practical need for a data model that encompasses the essential information researchers require to enhance explainability, fairness, and robustness in Computer Vision applications. We developed a data model that holds the potential to improve transparency and reproducibility within this field, speed up the research process, and facilitate comprehensive evaluations, whether quantitative or qualitative, of proposed methodologies. To refine and demonstrate the practicality of the data model, we have populated it with four existing datasets. Additionally, we have conducted two user studies to validate the model's usability. We found that participants were enthusiastic about using the data model. Some potential uses described by the participants were comparing models and datasets, accessing (niche) datasets and models, creating and exploring datasets, and having access to ground truth explanations. However, participants also had concerns about the data model, mainly with its usability being restricted to people with database knowledge and the richness of data in the database. Nonetheless, hope that this research constitutes the first step for data modelling for researchers in the field of Trustworthy AI.https://github.com/delftcrowd/CV_datamodel Code on GithubComputer Science | Data Science and Technolog
