Zhe Fu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Zhe Fu has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Xi Niu
University of North Carolina at Charlotte
6 shared publications - Ruhani Rahman
University of North Carolina at Charlotte
2 shared publications - Depeng Xu
University of North Carolina at Charlotte
1 shared publication Dongsong Zhang
Belk College of Business & School of Data Science, The University of North Carolina at Charlotte, Charlotte, NC, USA
1 shared publication- Guohou Shan
D’Amore-McKim School of Business, Northeastern University, Boston, Massachusetts 02115
1 shared publication Li Yu
Renmin University of China, Beijing, China
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
- 24 April 2025
- Published by Elsevier in Information & Management
- Vol. 62 (6)
Stylometric characteristics of code-switched offensive language in social media
- 24 April 2025
- Published by Elsevier in Information & Management
- Vol. 62 (6)
Open AccessStylometric characteristics of code-switched offensive language in social media
- 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.
- Open Access
- 24 February 2025
- Published by SAGE Publications in Production and Operations Management
- Vol. 34 (8) , 2448-2470
MT-GPD: A Multimodal Deep Transfer Learning Model Enhanced by Auxiliary Mechanisms for Cross-Domain Online Fake News Detection
The proliferation of fake news, more recently multimodal fake news, poses a significant threat to individuals, organizations, and society. While online social media platforms have employed automated methods to combat fake news, they face two notable challenges: the scarcity of labeled data and the diversity of news domains. To enhance the effectiveness and efficiency of online platforms in mitigating the spread of fake news, this study proposes MT-GPD (multimodal deep transfer learning with gating network, model patch, and domain classifier) for cross-domain fake news detection. MT-GPD integrates three novel design artifacts as auxiliary mechanisms for enhancing multimodal deep transfer learning, including a gating network that captures the relative importance of textual and visual components of individual news articles for dynamic fusion; a customized model patch that balances detection performance and computational efficiency; and a domain classifier that adapts multimodal representations to a target news domain. We evaluate the performance of MT-GPD using news datasets spanning four different domains. The results demonstrate the efficacy and robustness of MT-GPD, providing strong evidence for the impacts of the proposed auxiliary mechanisms on improving fake news detection performance.
- 24 February 2025
- Published by SAGE Publications in Production and Operations Management
- Vol. 34 (8) , 2448-2470
Open AccessMT-GPD: A Multimodal Deep Transfer Learning Model Enhanced by Auxiliary Mechanisms for Cross-Domain Online Fake News Detection
The proliferation of fake news, more recently multimodal fake news, poses a significant threat to individuals, organizations, and society. While online social media platforms have employed automated methods to combat fake news, they face two notable challenges: the scarcity of labeled data and the diversity of news domains. To enhance the effectiveness and efficiency of online platforms in mitigating the spread of fake news, this study proposes MT-GPD (multimodal deep transfer learning with gating network, model patch, and domain classifier) for cross-domain fake news detection. MT-GPD integrates three novel design artifacts as auxiliary mechanisms for enhancing multimodal deep transfer learning, including a gating network that captures the relative importance of textual and visual components of individual news articles for dynamic fusion; a customized model patch that balances detection performance and computational efficiency; and a domain classifier that adapts multimodal representations to a target news domain. We evaluate the performance of MT-GPD using news datasets spanning four different domains. The results demonstrate the efficacy and robustness of MT-GPD, providing strong evidence for the impacts of the proposed auxiliary mechanisms on improving fake news detection performance.
- Open Access
- 2 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 157-166
The Art of Asking: Prompting Large Language Models for Serendipity Recommendations
Serendipity means an unexpected but valuable discovery. Its elusive nature makes it susceptible to modeling. In this paper, we address the challenge of modeling serendipity in recommender systems using Large Language Models (LLMs), a recent breakthrough in AI technologies. We leveraged LLMs' prompting mechanisms to convert a problem of serendipity recommendations into a problem of formulating a prompt to elicit serendipity recommendations. The formulated prompt is called SerenPrompt. We designed three types of SerenPrompt: discrete with natural words, continuous with trainable tokens, and hybrid that combines the previous two types. In the meanwhile, for each type of SerenPrompt, we also designed two styles: direct and indirect, to investigate whether it is feasible to directly ask an LLM a question on whether an item is a serendipity, or we should breakdown the question into several sub-questions. Extensive experiments have demonstrated the effectiveness of SerenPrompt in generating serendipity recommendations, compared to the state-of-the-art models. The combination of the hybrid type and the indirect style achieves the best performance, with relatively low sacrifice to computational efficiency. The results demonstrate that natural words and virtual tokens, as building blocks of LLM prompts, each have their own advantages. The better performance of the indirect style speaks to the effectiveness of decomposing the direct question on serendipity.
- 2 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 157-166
Open AccessThe Art of Asking: Prompting Large Language Models for Serendipity Recommendations
Serendipity means an unexpected but valuable discovery. Its elusive nature makes it susceptible to modeling. In this paper, we address the challenge of modeling serendipity in recommender systems using Large Language Models (LLMs), a recent breakthrough in AI technologies. We leveraged LLMs' prompting mechanisms to convert a problem of serendipity recommendations into a problem of formulating a prompt to elicit serendipity recommendations. The formulated prompt is called SerenPrompt. We designed three types of SerenPrompt: discrete with natural words, continuous with trainable tokens, and hybrid that combines the previous two types. In the meanwhile, for each type of SerenPrompt, we also designed two styles: direct and indirect, to investigate whether it is feasible to directly ask an LLM a question on whether an item is a serendipity, or we should breakdown the question into several sub-questions. Extensive experiments have demonstrated the effectiveness of SerenPrompt in generating serendipity recommendations, compared to the state-of-the-art models. The combination of the hybrid type and the indirect style achieves the best performance, with relatively low sacrifice to computational efficiency. The results demonstrate that natural words and virtual tokens, as building blocks of LLM prompts, each have their own advantages. The better performance of the indirect style speaks to the effectiveness of decomposing the direct question on serendipity.
- Open Access
- 16 October 2023
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Knowledge Discovery From Data
- Vol. 18 (1) , 1-23
Modeling Users’ Curiosity in Recommender Systems
Today’s recommender systems are criticized for recommending items that are too obvious to arouse users’ interests. Therefore, the research community has advocated some “beyond accuracy” evaluation metrics such as novelty, diversity, and serendipity with the hope of promoting information discovery and sustaining users’ interests over a long period of time. While bringing in new perspectives, most of these evaluation metrics have not considered individual users’ differences in their capacity to experience those “beyond accuracy” items. Open-minded users may embrace a wider range of recommendations than conservative users. In this article, we proposed to use curiosity traits to capture such individual users’ differences. We developed a model to approximate an individual’s curiosity distribution over different stimulus levels. We used an item’s surprise level to estimate the stimulus level and whether such a level is in the range of the user’s appetite for stimulus, called Comfort Zone . We then proposed a recommender system framework that considers both user preference and their Comfort Zone where the curiosity is maximally aroused. Our framework differs from a typical recommender system in that it leverages human’s Comfort Zone for stimuli to promote engagement with the system. A series of evaluation experiments have been conducted to show that our framework is able to rank higher the items with not only high ratings but also high curiosity stimulation. The recommendation list generated by our algorithm has a higher potential of inspiring user curiosity compared to the state-of-the-art deep learning approaches. The personalization factor for assessing the surprise stimulus levels further helps the recommender model achieve smaller (better) inter-user similarity.
- 16 October 2023
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Knowledge Discovery From Data
- Vol. 18 (1) , 1-23
Open AccessModeling Users’ Curiosity in Recommender Systems
Today’s recommender systems are criticized for recommending items that are too obvious to arouse users’ interests. Therefore, the research community has advocated some “beyond accuracy” evaluation metrics such as novelty, diversity, and serendipity with the hope of promoting information discovery and sustaining users’ interests over a long period of time. While bringing in new perspectives, most of these evaluation metrics have not considered individual users’ differences in their capacity to experience those “beyond accuracy” items. Open-minded users may embrace a wider range of recommendations than conservative users. In this article, we proposed to use curiosity traits to capture such individual users’ differences. We developed a model to approximate an individual’s curiosity distribution over different stimulus levels. We used an item’s surprise level to estimate the stimulus level and whether such a level is in the range of the user’s appetite for stimulus, called Comfort Zone . We then proposed a recommender system framework that considers both user preference and their Comfort Zone where the curiosity is maximally aroused. Our framework differs from a typical recommender system in that it leverages human’s Comfort Zone for stimuli to promote engagement with the system. A series of evaluation experiments have been conducted to show that our framework is able to rank higher the items with not only high ratings but also high curiosity stimulation. The recommendation list generated by our algorithm has a higher potential of inspiring user curiosity compared to the state-of-the-art deep learning approaches. The personalization factor for assessing the surprise stimulus levels further helps the recommender model achieve smaller (better) inter-user similarity.
- Open Access
- 26 August 2023
- Published by Association for Computing Machinery (ACM) in ACM Computing Surveys
- Vol. 56 (1) , 1-26
Deep Learning Models for Serendipity Recommendations: A Survey and New Perspectives
Serendipitous recommendations have emerged as a compelling approach to deliver users with unexpected yet valuable information, contributing to heightened user satisfaction and engagement. This survey presents an investigation of the most recent research in serendipity recommenders, with a specific emphasis on deep learning recommendation models. We categorize these models into three types, distinguishing their integration of the serendipity objective across distinct stages: pre-processing, in-processing, and post-processing. Additionally, we provide a review and summary of the serendipity definition, available ground truth datasets, and evaluation experiments employed in the field. We propose three promising avenues for future exploration: (1) leveraging user reviews to identify and explore serendipity, (2) employing reinforcement learning to construct a model for discerning appropriate timing for serendipitous recommendations, and (3) utilizing cross-domain learning to enhance serendipitous recommendations. With this review, we aim to cultivate a deeper understanding of serendipity in recommender systems and inspire further advancements in this domain.
- 26 August 2023
- Published by Association for Computing Machinery (ACM) in ACM Computing Surveys
- Vol. 56 (1) , 1-26
Open AccessDeep Learning Models for Serendipity Recommendations: A Survey and New Perspectives
Serendipitous recommendations have emerged as a compelling approach to deliver users with unexpected yet valuable information, contributing to heightened user satisfaction and engagement. This survey presents an investigation of the most recent research in serendipity recommenders, with a specific emphasis on deep learning recommendation models. We categorize these models into three types, distinguishing their integration of the serendipity objective across distinct stages: pre-processing, in-processing, and post-processing. Additionally, we provide a review and summary of the serendipity definition, available ground truth datasets, and evaluation experiments employed in the field. We propose three promising avenues for future exploration: (1) leveraging user reviews to identify and explore serendipity, (2) employing reinforcement learning to construct a model for discerning appropriate timing for serendipitous recommendations, and (3) utilizing cross-domain learning to enhance serendipitous recommendations. With this review, we aim to cultivate a deeper understanding of serendipity in recommender systems and inspire further advancements in this domain.
- Open Access
- 18 July 2023
- Published by Association for Computing Machinery (ACM)
- p. 739-748
Wisdom of Crowds and Fine-Grained Learning for Serendipity Recommendations
- 18 July 2023
- Published by Association for Computing Machinery (ACM)
- p. 739-748
Open AccessWisdom of Crowds and Fine-Grained Learning for Serendipity Recommendations
- Open Access
- 8 January 2022
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Knowledge Discovery From Data
- Vol. 16 (4) , 1-22
TRACE: Travel Reinforcement Recommendation Based on Location-Aware Context Extraction
As the popularity of online travel platforms increases, users tend to make ad-hoc decisions on places to visit rather than preparing the detailed tour plans in advance. Under the situation of timeliness and uncertainty of users’ demand, how to integrate real-time context into dynamic and personalized recommendations have become a key issue in travel recommender system. In this article, by integrating the users’ historical preferences and real-time context, a location-aware recommender system called TRACE ( T ravel R einforcement Recommendations Based on Location- A ware C ontext E xtraction) is proposed. It captures users’ features based on location-aware context learning model, and makes dynamic recommendations based on reinforcement learning. Specifically, this research: (1) designs a travel reinforcing recommender system based on an Actor-Critic framework, which can dynamically track the user preference shifts and optimize the recommender system performance; (2) proposes a location-aware context learning model, which aims at extracting user context from real-time location and then calculating the impacts of nearby attractions on users’ preferences; and (3) conducts both offline and online experiments. Our proposed model achieves the best performance in both of the two experiments, which demonstrates that tracking the users’ preference shifts based on real-time location is valuable for improving the recommendation results.
- 8 January 2022
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Knowledge Discovery From Data
- Vol. 16 (4) , 1-22
Open AccessTRACE: Travel Reinforcement Recommendation Based on Location-Aware Context Extraction
As the popularity of online travel platforms increases, users tend to make ad-hoc decisions on places to visit rather than preparing the detailed tour plans in advance. Under the situation of timeliness and uncertainty of users’ demand, how to integrate real-time context into dynamic and personalized recommendations have become a key issue in travel recommender system. In this article, by integrating the users’ historical preferences and real-time context, a location-aware recommender system called TRACE ( T ravel R einforcement Recommendations Based on Location- A ware C ontext E xtraction) is proposed. It captures users’ features based on location-aware context learning model, and makes dynamic recommendations based on reinforcement learning. Specifically, this research: (1) designs a travel reinforcing recommender system based on an Actor-Critic framework, which can dynamically track the user preference shifts and optimize the recommender system performance; (2) proposes a location-aware context learning model, which aims at extracting user context from real-time location and then calculating the impacts of nearby attractions on users’ preferences; and (3) conducts both offline and online experiments. Our proposed model achieves the best performance in both of the two experiments, which demonstrates that tracking the users’ preference shifts based on real-time location is valuable for improving the recommendation results.