Zheng Zhang

Dr. Zheng Zhang

orcid

Research Focus & Metrics

Representative Publications

View All Publications
21 January 2025 in IEEE Access
Verified
Journal Article
Open access
AAAish Albladi
Aish Albladi
MIMinarul Islam
Minarul Islam
ADAmit Das
Amit Das
MBMaryam Bigonah
Maryam Bigonah
ZZZheng Zhang
Zheng Zhang
CSCheryl Seals
Cheryl Seals
The widespread use of social media and other online platforms has facilitated unprecedented communication and information exchange. However, it has also led to the spread of hate speech and poses serious challenges to societal harmony as well as individual well-being. Traditional methods for detecting hate speech, such as keyword matching, rule-based systems, and machine learning algorithms, often struggle to capture the subtle and context-dependent nature of hateful content. This paper provides a comprehensive review of the application of large language models (LLMs) like GPT-3, BERT, and their successors in hate speech detection. We analyze the evolution of LLMs in natural language processing and examine their strengths and limitations in identifying hate speech. Additionally, we address the significant challenges and explore how LLMs method can affect the accuracy and fairness of hate speech detection systems. By synthesizing recent research, this review aims to offer a holistic understanding of the current state-of-the-art methods in hate speech detection utilizing LLMs and to suggest directions for future research that could enhance the efficacy and equity of these systems.

5 April 2023 in IEEE Transactions on Neural Networks and Learning Systems
Verified
Journal Article
Closed access
ZZZheng Zhang
Zheng Zhang
LYLevent Yilmaz
Levent Yilmaz
BLBo Liu
Bo Liu
Despite recent advances in modern machine learning algorithms, the opaqueness of their underlying mechanisms continues to be an obstacle in adoption. To instill confidence and trust in artificial intelligence (AI) systems, explainable AI (XAI) has emerged as a response to improve modern machine learning algorithms’ explainability. Inductive logic programming (ILP), a subfield of symbolic AI, plays a promising role in generating interpretable explanations because of its intuitive logic-driven framework. ILP effectively leverages abductive reasoning to generate explainable first-order clausal theories from examples and background knowledge. However, several challenges in developing methods inspired by ILP need to be addressed for their successful application in practice. For example, the existing ILP systems often have a vast solution space, and the induced solutions are very sensitive to noises and disturbances. This survey paper summarizes the recent advances in ILP and a discussion of statistical relational learning (SRL) and neural-symbolic algorithms, which offer synergistic views to ILP. Following a critical review of the recent advances, we delineate observed challenges and highlight potential avenues of further ILP-motivated research toward developing self-explanatory AI systems.

14 June 2025 in Peer-to-Peer Networking and Applications
Verified
Journal Article
Closed access
PDF file
ZZZheng Zhang
Zheng Zhang
ADAmit Das
Amit Das
GHGuan Huang
Guan Huang
SBSanjeev Baskiyar
Sanjeev Baskiyar

27 June 2025 in IEEE Access
Verified
Journal Article
Open access
YBYida Bao
Yida Bao
ZZZheng Zhang
Zheng Zhang
MAMohammad Arifuzzaman
Mohammad Arifuzzaman
Tran Duc LeTran Duc Le
Tran Duc Le
QLQi Li
Qi Li
JYJiafeng Ye
Jiafeng Ye
Recognizing Shanghai dialect text is crucial for preserving local dialects, yet research on its automatic distinction from Standard Mandarin remains limited. We construct a carefully balanced dataset specifically for the task of Shanghai dialect recognition and propose a two-stage approach for automatic language classification. In the first stage, we employ Jieba tokenization to retain dialect-specific lexical nuances, ensuring essential semantic and syntactic distinctions are captured. Next, we independently train both a BERT-Chinese-Based classifier and a traditional Support Vector Machine classifier for dialect recognition. The BERT model leverages powerful contextual representations to capture subtle differences between Shanghai dialect and Standard Mandarin, while the Support Vector Machine serves as a conventional baseline. Extensive experiments comparing the two approaches revealed that, although the Support Vector Machine can adequately perform the classification task, the BERT-Based classifier achieves significantly higher accuracy and is more sensitive to the nuanced linguistic features of the dialect. Further analysis through attention visualization reveals how the model specifically attends to unique dialectal features, highlighting distinctive lexical and structural differences between Shanghai dialect and Mandarin text. To the best of our knowledge, this study is the first to apply NLP techniques for language classification between Shanghai dialect and Standard Mandarin, emphasizing the potential for automated dialect recognition as an effective method for dialect documentation and preservation.

20 April 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
YBYida Bao
Yida Bao
PGPhilippe Gaillard
Philippe Gaillard
WYWei Yao
Wei Yao
ZZZheng Zhang
Zheng Zhang
RWRui Wang
Rui Wang
Dynamic Time Warping (DTW) is a useful method for comparing time-series signals when timing shifts or delays are present. Common evaluation metrics such as Mean Squared Error (MSE) focus on point-by-point differences and often fail to reflect similarity in overall shape or timing. As a result, models that visually match the behavior of a biological signal may still receive poor numerical scores. In this paper, we implement DTW in SAS 9.4 using PROC IML and provide a practical way to compute dynamic distances between time-series signals directly within SAS. Using both simple examples and real ECG data, we show that DTW captures waveform similarity more effectively than traditional error-based measures when temporal misalignment exists. This work provides SAS users with a straightforward and interpretable tool for evaluating time-dependent data.

26 October 2026 in Multimedia
Verified
Journal Article
Open access
PDF file
HMHassan Mistareehi
Hassan Mistareehi
MNMd Sultan Al Nahian
Md Sultan Al Nahian
ZZZheng Zhang
Zheng Zhang
The ubiquitous generation of multimedia data—images, video, and audio—by smartphones, surveillance networks, and wearable sensors has created unprecedented analytical opportunities while amplifying privacy risks. Unlike structured tabular data, identity in multimedia is redundantly embedded in pixels, waveforms, and metadata, rendering traditional de-identification methods brittle. Furthermore, modern deep learning models can re-identify individuals even after conventional redaction. This paper organizes privacy-preserving multimedia techniques across the data lifecycle: collection, analysis, and dissemination. We examine the principal technical families, including visual/acoustic obfuscation, differential privacy, federated and split learning, homomorphic encryption, secure multi-party computation, generative adversarial approaches for synthetic data, and blockchain-based access control. We synthesize the state of the art regarding the privacy–utility trade-off, summarize representative attacks (re-identification, membership inference, model inversion, and gradient leakage), and present a comparative taxonomy across modalities, lifecycle stages, and computational costs. We further offer a cross-cutting discussion identifying four themes—the insufficiency of any single mechanism in isolation, an under-addressed adaptive-adversary problem, an unresolved gap between formal privacy guarantees and perceptual/task utility, and uneven coverage across modalities—and conclude by identifying open challenges, including scalable encrypted video processing, standardized privacy–utility metrics, regulatory harmonization, and robustness against adaptive adversaries, that define the future research agenda.