Ling Lin

Violence detection algorithms hold significant potential for security systems and online video content moderation. While current video analysis methods demonstrate effectiveness in identifying anomalous behaviors within controlled environments, two critical limitations persist: (1) Insufficient generalization capacity for novel application scenarios, and (2) Discrepancy between balanced benchmark datasets and real-world imbalanced data distributions, where violent incidents exhibit extreme rarity (typically <1%) compared to routine activities. To address these challenges, we first construct the Open-World Fights (OWF) dataset, featuring a realistic 1:100 violence-to-normal behavior ratio. We then propose GVD, a generalized violence detection framework combining universal recognition capabilities with long-tail distribution adaptation. The architecture incorporates two novel components: a Semantic Differentiation Module (SDM) that mitigates background interference through foreground-background decoupling, and an Orthogonal Supervision Module (OSM) that optimizes cross-modal alignment via feature orthogonality constraints. Extensive evaluations demonstrate GVD’s competitive performance against contemporary baselines across multiple benchmarks, particularly in out-of-distribution and imbalanced data settings. Practical experiments suggest the framework’s robustness in specific real-world deployments, such as continuous monitoring with the OWF dataset, though scalability challenges remain.

Introduction: In the 2020s, particularly following 2022, the Chinese government introduced a series of initiatives to foster the development of the prepared dishes sector, accompanied by substantial investments from industrial capital. Consequently, China’s prepared dishes industry has experienced rapid growth. Nevertheless, this swift expansion has elicited varied public opinions, particularly concerning the potential health effects of prepared dishes. Therefore, this study aims to gather and analyze comments from social media on prepared dishes using machine learning techniques. The objective is to ascertain the perspectives of the Chinese populace on the health implications of consuming prepared dishes.Methods: Social media comments, characterized by their broad distribution, objectivity, and timeliness, served as the primary data source for this study. Initially, the data underwent preprocessing to ensure its suitability for analysis. Subsequent steps in this study involved conducting sentiment analysis and employing the BERTopic model for topic clustering. These methods aimed to identify the principal concerns of the public regarding the impact of prepared dishes on health. The final phase of the study involved a comparative analysis of changes in public sentiment and thematic focus across different time frames. This approach provides a dynamic view of evolving public perceptions related to the health implications of prepared dishes.Results: This study analyzed over 600,000 comments gathered from various social media platforms from mid-July 2022 to the end of March 2024. Following data preprocessing, 200,993 comments were assessed for sentiment, revealing that more than 64% exhibited negative emotions. Subsequent topic clustering using the BERTopic model identified that 11 of the top 50 topics were related to public health concerns. These topics primarily scrutinized the safety of prepared dish production processes, raw materials, packaging materials, and additives. Moreover, significant public’s interest was in the right to informed consumption across different contexts. Notably, the most pronounced public opposition emerged regarding introducing prepared dishes into primary and secondary school canteens, with criticisms directed at the negligence of educational authorities and the ethics of manufacturers. Additionally, there were strong recommendations for media organizations to play a more active role in monitoring public opinion and for government agencies to enhance regulatory oversight.Conclusion: The findings of this study indicate that more than half of the Chinese public maintain a negative perception towards prepared dishes, particularly concerning about health implications. Chinese individuals display considerable sensitivity and intense reactions to news and events related to prepared dishes. Consequently, the study recommends that manufacturers directly address public psychological perceptions, proactively enhance production processes and service quality, and increase transparency in public communications to improve corporate image and people acceptance of prepared dishes. Additionally, supervisory and regulatory efforts must be intensified by media organizations and governmental bodies, fostering the healthy development of the prepared food industry in China.

Online gaming constitutes an indispensable facet of China’s digital economy, catalyzing consumer discussions on social media platforms. This study employs a comprehensive natural language processing framework, encompassing topic mining, multi-label classification, and sentiment analysis, to evaluate consumers’ psychological perceptions of the risks associated with online games through social media comments. This study identifies 11 distinct perceived risk topics, including “Excessive Temptation”, “Entry Regulation”, and “Culture Implantation”. Numerous comments encompass multiple topics, each infused with diverse emotional inclinations, thus unveiling disparate consumer perspectives. These findings underscore the critical significance of addressing potential perceived risks and mitigating negative consumer emotions for enterprises operating within online gaming. Such measures are pivotal to maintaining a brand image, business reputation, and enduring growth. Furthermore, this study extends valuable insights to regulatory bodies, contributing to enhancing administrative efficiency, safeguarding consumer rights, and fostering a robust and sustainable trajectory within China’s online gaming market.

In recent years, as people’s awareness of energy conservation, environmental protection, and sustainable development has increased, discussions related to electric vehicles (EVs) have aroused public debate on social media. At some point, most consumers face the possible risks of EVs—a critical psychological perception that invariably affects sales of EVs in the consumption market. This paper chooses to deconstruct customers’ perceived risk from third-party comment data in social media, which has better coverage and objectivity than questionnaire surveys. In order to analyze a large amount of unstructured text comment data, the natural language processing (NLP) method based on machine learning was applied in this paper. The measurement results show 15 abstracts in five consumer perceived risks to EVs. Among them, the largest number of comments is that of “Technology Maturity” (A13) which reached 25,329, and which belongs to the “Performance Risk” (PR1) dimension, indicating that customers are most concerned about the performance risk of EVs. Then, in the “Social Risk” (PR5) dimension, the abstract “Social Needs” (A51) received only 3224 comments and “Preference and Trust Rank” (A52) reached 22,324 comments; this noticeable gap indicated the changes in how consumers perceived EVs social risks. Moreover, each dimension’s emotion analysis results showed that negative emotions are more than 40%, exceeding neutral or positive emotions. Importantly, customers have the strongest negative emotions about the “Time Risk” (PR4), accounting for 54%. On a finer scale, the top three negative emotions are “Charging Time” (A42), “EV Charging Facilities” (A41), and “Maintenance of Value” (A33). Another interesting result is that “Social Needs” (A51)’s positive emotional comments were larger than negative emotional comments. The paper provides substantial evidence for perceived risk theory research by new data and methods. It can provide a novel tool for multi-dimensional and fine-granular capture customers’ perceived risks and negative emotions. Thus, it has the potential to help government and enterprises to adjust promotional strategies in a timely manner to reduce higher perceived risks and emotions, accelerating the sustainable development of EVs’ consumption market in China.