20 January 2021 in IEEE Access
The purpose of this article is to analyse the emotional evolution of the netizens in reaction to the events of the Anti-ELAB (Anti-Extradition Law Amendment Bill) movement in Hong Kong. We attempt to investigate evolving laws of large-scale Internet public opinion events and provide relevant agencies with a theoretical basis for a public opinion response mechanism. On the basis of improving the Bidirectional Encoder Representations from Transformers (BERT) pre-training task, we add in-depth pre-training tasks, and based on the optimisation results of the LDA topic embedding, we integrate deeply with the LDA model to dynamically present the fine-grained public sentiment of the event. Through the collection of large-scale text data related to the Anti-ELAB Movement from a well-known forum in Hong Kong, a BERT-LDA hybrid model for large-scale network public opinion analysis is constructed in a complex context. Through empirical analysis, we calculate and reveal the emotional change process of netizens and opinion leaders in the three transition stages of the Anti-ELAB Movement with the evolution of the topic word as the core by visualisation. We also analyse the emotional distribution and evolution trend of public opinion under the `text topic', and deeply analyse the character and role of opinion leaders in Anti-ELAB public opinion events. The improved BERT-LDA model or sentiment classification AUC value exceeds 99.6% in the sentiment classification task for the Anti-ELAB Movement.
31 October 2024 in PLOS ONE
Postgraduate students face various academic, personal, and social stressors that increase their risk of anxiety, depression, and suicide. Identifying cost-effective methods of detecting and intervening before stress turns into severe problems is crucial. However, existing stress detection methods typically rely on psychological scales or devices, which can be complex and expensive. Therefore, we propose a BERT-fused model for rapidly and automatically detecting postgraduate students’ psychological stress via social media. First, we construct an improved BERT-LDA feature extraction algorithm to extract group stress features from large-scale and complex social media data. Then, we integrate the BiLSTM-CRF named entity recognition model to construct a multi-dimensional psychological stress profile and analyze the fine-grained feature representation under the fusion of multi-dimensional features. Experimental results demonstrate that the proposed model outperforms traditional models such as BiLSTM, achieving an accuracy of 92.55%, a recall of 93.47%, and an F1-score of 92.18%, with F1-scores exceeding 89% for all three types of entities. This research provides both theoretical and practical foundations for universities or institutions to conduct fine-grained perception and intervention for postgraduate students’ psychological stress.