Ruhani Rahman
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Ruhani Rahman has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Xi Niu
University of North Carolina at Charlotte
2 shared publications - Zhe Fu
Software and Information Systems, The University of North Carolina at Charlotte, Charlotte, United States
2 shared publications - Depeng Xu
University of North Carolina at Charlotte
1 shared publication - Shi Chen
University of North Carolina at Charlotte
1 shared publication Shuhua Yin
1 shared publication- Xiangcheng Wu
Software and Information Systems, The University of North Carolina at Charlotte, Charlotte, United States
1 shared publication
Latest Publications
- 25 June 2026
- Published by Springer Nature
- p. 164-176
Reducing Disparity in Aleatoric Uncertainty for Age Estimation from Facial Images
- 25 June 2026
- Published by Springer Nature
- p. 164-176
Reducing Disparity in Aleatoric Uncertainty for Age Estimation from Facial Images
- Open Access
- 26 August 2025
- Published by JMIR Publications Inc. in JMIR Preprints
Assess Topic Consistency Between CDC Messaging and Public Responses during the COVID-19 Pandemic: Infoveillance Study (Preprint)
BACKGROUND During the COVID-19 pandemic, the U.S. Centers for Disease Control and Prevention (CDC) actively used social media to disseminate information. However, traditional engagement metrics (eg, likes, shares, replies) do not capture whether public responses are thematically consistent with official communications, leaving a critical gap in evaluating the effectiveness of digital health messaging. OBJECTIVE This study aimed to develop and evaluate a novel infoveillance framework that quantifies the dynamic consistency between CDC COVID-19 communications and public responses on social media. METHODS We collected 17,524 CDC posts and 26,812 public replies from Twitter/X between May 14, 2020, and November 10, 2022. Using the Biterm Topic Model (BTM), we extracted latent topics from CDC communications and public responses. We then developed a topic consistency scoring system that combined semantic similarity (cosine similarity of topic-word distributions) with daily public topic prominence (weights of response topics per day). The resulting weighted similarity scores generated a time series to assess topic alignment dynamically. RESULTS BTM identified 9 CDC topics and 8 public response topics. Alignment varied across topics and over time. CDC topics related to epidemic outcomes (eg, hospitalization, death) and high-risk populations consistently showed high alignment with public responses, especially during Delta and Omicron variant phases. In contrast, topics on non-pharmaceutical interventions exhibited lower weighted similarity, indicating weaker public engagement despite moderate semantic similarity. The weighted similarity metric outperformed conventional semantic similarity measures by accounting for daily shifts in public attention. CONCLUSIONS This study introduces a flexible and modular framework for public health infoveillance that quantifies dynamic alignment between official health communications and public discourse. Findings highlight which CDC messages were most and least aligned with public concerns, offering evidence to improve communication strategies during future health emergencies. The framework is generalizable and can integrate alternative topic models, similarity measures, and social media platforms.
- 26 August 2025
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessAssess Topic Consistency Between CDC Messaging and Public Responses during the COVID-19 Pandemic: Infoveillance Study (Preprint)
BACKGROUND During the COVID-19 pandemic, the U.S. Centers for Disease Control and Prevention (CDC) actively used social media to disseminate information. However, traditional engagement metrics (eg, likes, shares, replies) do not capture whether public responses are thematically consistent with official communications, leaving a critical gap in evaluating the effectiveness of digital health messaging. OBJECTIVE This study aimed to develop and evaluate a novel infoveillance framework that quantifies the dynamic consistency between CDC COVID-19 communications and public responses on social media. METHODS We collected 17,524 CDC posts and 26,812 public replies from Twitter/X between May 14, 2020, and November 10, 2022. Using the Biterm Topic Model (BTM), we extracted latent topics from CDC communications and public responses. We then developed a topic consistency scoring system that combined semantic similarity (cosine similarity of topic-word distributions) with daily public topic prominence (weights of response topics per day). The resulting weighted similarity scores generated a time series to assess topic alignment dynamically. RESULTS BTM identified 9 CDC topics and 8 public response topics. Alignment varied across topics and over time. CDC topics related to epidemic outcomes (eg, hospitalization, death) and high-risk populations consistently showed high alignment with public responses, especially during Delta and Omicron variant phases. In contrast, topics on non-pharmaceutical interventions exhibited lower weighted similarity, indicating weaker public engagement despite moderate semantic similarity. The weighted similarity metric outperformed conventional semantic similarity measures by accounting for daily shifts in public attention. CONCLUSIONS This study introduces a flexible and modular framework for public health infoveillance that quantifies dynamic alignment between official health communications and public discourse. Findings highlight which CDC messages were most and least aligned with public concerns, offering evidence to improve communication strategies during future health emergencies. The framework is generalizable and can integrate alternative topic models, similarity measures, and social media platforms.
- Open Access
- 21 March 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Recommender Systems
- Vol. 3 (3) , 1-21
A Deep Learning Model for Cross-Domain Serendipity Recommendations
Serendipity means unexpected discoveries that are valuable, with positive outcomes ranging from personal benefits to scientific breakthroughs. This study proposes a cross-domain recommendation model, called SerenCDR , to model serendipity. SerenCDR leverages the knowledge beyond one domain as well as mitigates the inherent data sparsity problem in serendipity recommendations. The novelty of SerenCDR lies in the fact that it is the first deep learning based cross-domain model for a serendipity task. More importantly, it does not rely on any overlapping users or overlapping items across different domains, which especially fits for the task of recommending serendipity, because serendipity in a single domain tends to be sparse; finding overlapping users or overlapping items in other domains is nearly impossible. To train and test SerenCDR , we have collected a two-domain ground truth dataset on serendipity, called SerenCDRLens . In addition, since we found that serendipity is sparse in SerenCDRLens , we designed an auxiliary loss function to supplement the main loss function to enhance serendipity learning. Through a series of experiments, we have harvested positive performance in recommending serendipity, empowering users with increased chances of bumping into unexpected but valuable discoveries.
- 21 March 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Recommender Systems
- Vol. 3 (3) , 1-21
Open AccessA Deep Learning Model for Cross-Domain Serendipity Recommendations
Serendipity means unexpected discoveries that are valuable, with positive outcomes ranging from personal benefits to scientific breakthroughs. This study proposes a cross-domain recommendation model, called SerenCDR , to model serendipity. SerenCDR leverages the knowledge beyond one domain as well as mitigates the inherent data sparsity problem in serendipity recommendations. The novelty of SerenCDR lies in the fact that it is the first deep learning based cross-domain model for a serendipity task. More importantly, it does not rely on any overlapping users or overlapping items across different domains, which especially fits for the task of recommending serendipity, because serendipity in a single domain tends to be sparse; finding overlapping users or overlapping items in other domains is nearly impossible. To train and test SerenCDR , we have collected a two-domain ground truth dataset on serendipity, called SerenCDRLens . In addition, since we found that serendipity is sparse in SerenCDRLens , we designed an auxiliary loss function to supplement the main loss function to enhance serendipity learning. Through a series of experiments, we have harvested positive performance in recommending serendipity, empowering users with increased chances of bumping into unexpected but valuable discoveries.