Peng Zhang

Peng Zhang


Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
6 June 2025 in PLOS ONE
Verified
Journal Article
Open access
PDF file
PZPeng Zhang
Peng Zhang
YDYinghui Du
Yinghui Du
QWQilei Wang
Qilei Wang
JZJiyang Zhang
Jiyang Zhang
RQRuiqing Qin
Ruiqing Qin
Accurately identifying social bot accounts is the key to preventing the use of artificial intelligence technology to forge social accounts, which can interfere with public opinion and thus cause public opinion crises. However, at present, relying only on manual identification of bot accounts has the challenges of low efficiency, high cost, and low accuracy, while existing research on batch identification of social bots lacks research on the system of behavioural characteristics of social bots, and thus lacks the construction of a model for the analysis of the behavioural characteristics of social bots. In this paper, we propose a diverse set of behavioural features for social robots based on the differences between the behavioural features of social robot accounts and normal users. The feature selection method based on OOB estimation is chosen for excluding redundant features in the constructed feature set; meanwhile, Random Forest, as a combined classification method, overcomes the problem of limitations of decision boundaries when classifying with a single decision tree, and has the characteristics of high accuracy, fast speed and stable performance. Through experiments, this paper applies it to the construction of social robot recognition model for detecting robot accounts in social platforms. The experiments prove that the effective indicators screened by the feature selection method based on OOB estimation can help improve the stability of the model. Specifically, the filtered features contribute about 20% more to the model accuracy and F1 score than other features. The social robot recognition model constructed based on random forest has higher accuracy and stability compared to the decision tree model and neural network model. Specifically, the accuracy rate is about 5% higher than other models, and other indicators are also better than other models. The experimental results show that the feature selection method based on OOB estimation and the random forest model show excellent performance in the experiments of social robot recognition, which can meet the requirements of the actual social robot recognition research and can be applied to the practical scenarios of robot account detection on social platforms.

16 March 2026 in ICST Transactions on Scalable Information Systems
Verified
Journal Article
Open access
PDF file
HWHongliang Wang
Hongliang Wang
SZShoumin Zhang
Shoumin Zhang
NLNa Li
Na Li
JLJing Liu
Jing Liu
PZPeng Zhang
Peng Zhang
INTRODUCTION: The Internet's features of transcending time and space and anonymity have fostered more rampant and covert online violent speech. Thus, accurate and effective management of online public opinion is of great significance. In recent years, scholars both domestically and internationally have conducted extensive research on online violent speech detection, but current challenges include extracting semantics from diverse and implicit expressions in Chinese online violent short texts. OBJECTIVES: This paper aims to propose the EBLA model for online violent speech detection, based on the ERNIE knowledge-enhanced semantic understanding pre-training model and the BiLSTM-Attention network, to precisely identify relevant textual semantic information and provide an effective method for online content moderators. METHODS: The model is trained using publicly available Chinese datasets related to online violence. It enhances deep, sentence-level feature extraction by integrating an attention mechanism into the BiLSTM layer on top of the ERNIE pre-training model. The model consists of vector transformation, deep text feature extraction, and text classification prediction phases. RESULTS: Results show that the precision of this model in identifying Chinese online violence tasks surpasses the BERT pre-training model by 3.7% and outperforms the BiLSTM combined with the attention mechanism by 13.84%. Empirical studies on additional datasets confirm the model's robustness and transferability. CONCLUSION: The EBLA model provides a strong basis for online violent speech detection, though it has limitations such as not accounting for identity bias or dynamic speech nature. Future improvements will focus on multimodal analysis and dynamic monitoring capabilities.