A Privacy-Preserving Semi-Decentralized Personalized Recommendation System DOI
Carson K. Leung,

Evan W.R. Madill,

Qi Wen

et al.

2021 IEEE International Conference on Big Data (Big Data), Journal Year: 2023, Volume and Issue: unknown, P. 1336 - 1343

Published: Dec. 15, 2023

In the present era of big data, recommendation systems play a crucial role in our daily lives by assisting us making quicker and more informed decisions from vast array choices. The concept personalized recommendations has gained widespread popularity, offering suggestions based on user profiles, preferences, and/or interests. Although many existing centralize data for recommendations, revelation sensitive poses privacy concern, as research indicates potential to de-identify anonymous users. For instance, information such political views or sexual orientations can be inferred seemingly non-sensitive like product review ratings. this paper, we privacy-preserving system named P2RecSys address these issues. Our takes semi-decentralized approach treating each node network an agent. Data are distributed agent within trusted networks, service provider only collects obfuscated agents using differential-privacy mechanism. Consequently, either safeguarded local networks outside networks. final is then generated combining with global provider. emphasis allows highly recommendations. Evaluation results demonstrate that achieves high accuracy while effectively safeguarding privacy.

Language: Английский

A Data Fusion Framework for Multi-Domain Morality Learning DOI Open Access
Siyi Guo,

Negar Mokhberian,

Kristina Lerman

et al.

Proceedings of the International AAAI Conference on Web and Social Media, Journal Year: 2023, Volume and Issue: 17, P. 281 - 291

Published: June 2, 2023

Language models can be trained to recognize the moral sentiment of text, creating new opportunities study role morality in human life. As interest language and has grown, several ground truth datasets with annotations have been released. However, these vary method data collection, domain, topics, instructions for annotators, etc. Simply aggregating such heterogeneous during training yield that fail generalize well. We describe a fusion framework on multiple improve performance generalizability. The model uses domain adversarial align feature space weighted loss function deal label shift. show proposed achieves state-of-the-art different compared prior works inference.

Language: Английский

Citations

13
When Infodemic Meets Epidemic: Systematic Literature Review DOI Creative Commons
Chaimae Asaad, Imane Khaouja, Mounir Ghogho

et al.

JMIR Public Health and Surveillance, Journal Year: 2025, Volume and Issue: 11, P. e55642 - e55642

Published: Feb. 3, 2025

Epidemics and outbreaks present arduous challenges, requiring both individual communal efforts. The significant medical, emotional, financial burden associated with epidemics creates feelings of distrust, fear, loss control, making vulnerable populations prone to exploitation manipulation through misinformation, rumors, conspiracies. use social media sites has increased in the last decade. As a result, amounts public data can be leveraged for biosurveillance. Social also provide platform quickly efficiently reach sizable percentage population; therefore, they have potential role various aspects epidemic mitigation. This systematic literature review aimed methodical overview integration 3 epidemic-related contexts: monitoring, misinformation detection, relationship mental health. aim is understand how been used these contexts, which gaps need further research Three questions, related health, were conceptualized this review. In first PRISMA (Preferred Reporting Items Systematic Reviews Meta-Analyses) stage, 13,522 publications collected from several digital libraries (PubMed, IEEE Xplore, ScienceDirect, SpringerLink, MDPI, ACM, ACL) gray sources (arXiv ProQuest), spanning 2010 2022. A total 242 (1.79%) papers selected inclusion synthesized identify themes, methods, studied, used. Five main themes identified literature, as follows: forecasting surveillance, opinion understanding, fake news identification characterization, health assessment, association psychological outcomes. found an efficient tool gauge response, monitor discourse, misleading news, estimate toll epidemics. Findings uncovered more robust applications lessons learned "postmortem documentation." vast gap exists between retrospective analysis management result prospective studies. Harnessing full tasks requires streamlining results forecasting, all while keeping abreast implications. Proactive prevention thus become vital curtailment containment.

Language: Английский

Citations

0
Can LLMs Assist Annotators in Identifying Morality Frames? - Case Study on Vaccination Debate on Social Media DOI
Tunazzina Islam, Dan Goldwasser

Published: May 19, 2025

Language: Английский

Citations

0
Weakly Supervised Learning for Analyzing Political Campaigns on Facebook DOI Open Access
Tunazzina Islam, Shamik Roy, Dan Goldwasser

et al.

Proceedings of the International AAAI Conference on Web and Social Media, Journal Year: 2023, Volume and Issue: 17, P. 411 - 422

Published: June 2, 2023

Social media platforms are currently the main channel for political messaging, allowing politicians to target specific demographics and adapt based on their reactions. However, making this communication transparent is challenging, as messaging tightly coupled with its intended audience often echoed by multiple stakeholders interested in advancing policies. Our goal paper take a first step towards understanding these highly decentralized settings. We propose weakly supervised approach identify stance issue of ads Facebook analyze how campaigns use some kind demographic targeting location, gender, or age. Furthermore, we temporal dynamics election polls.

Language: Английский

Citations

6
Analysis of Climate Campaigns on Social Media using Bayesian Model Averaging DOI Creative Commons
Tunazzina Islam, Ruqi Zhang, Dan Goldwasser

et al.

Published: Aug. 8, 2023

Climate change is the defining issue of our time, and we are at a moment. Various interest groups, social movement organizations, individuals engage in collective action on this media. In addition, advocacy campaigns media often arise response to ongoing societal concerns, especially those faced by energy industries. Our goal paper analyze how industries, their group, climate group use influence narrative change. work, propose minimally supervised model soup [57] approach combined with messaging themes identify stances ads Facebook. Finally, release stance dataset, model, set related for future work opinion mining automatic detection stances.

Language: Английский

Citations

2
A Privacy-Preserving Semi-Decentralized Personalized Recommendation System DOI
Carson K. Leung,

Evan W.R. Madill,

Qi Wen

et al.

2021 IEEE International Conference on Big Data (Big Data), Journal Year: 2023, Volume and Issue: unknown, P. 1336 - 1343

Published: Dec. 15, 2023

In the present era of big data, recommendation systems play a crucial role in our daily lives by assisting us making quicker and more informed decisions from vast array choices. The concept personalized recommendations has gained widespread popularity, offering suggestions based on user profiles, preferences, and/or interests. Although many existing centralize data for recommendations, revelation sensitive poses privacy concern, as research indicates potential to de-identify anonymous users. For instance, information such political views or sexual orientations can be inferred seemingly non-sensitive like product review ratings. this paper, we privacy-preserving system named P2RecSys address these issues. Our takes semi-decentralized approach treating each node network an agent. Data are distributed agent within trusted networks, service provider only collects obfuscated agents using differential-privacy mechanism. Consequently, either safeguarded local networks outside networks. final is then generated combining with global provider. emphasis allows highly recommendations. Evaluation results demonstrate that achieves high accuracy while effectively safeguarding privacy.

Language: Английский

Citations

2