| Sudden public safety incidents usually attract a high level of attention from the whole society,also evolve in both cyberspace and real society.This phenomenon increases the difficulty of government emergency management and presents an unprecedented challenge to the goal of social safety governance.In response to the more diverse and complex online interactions in the emerging media,the reactive approach to governance after the information release may increase the cost of negative social costs.Driven by the need of analyzing the social security situation with big data facing uncertainties and unknown risks,this research returns to the governance perspective of the social subject"people" as an important internal motivation to solve the social governance dilemma.By reconstructing the transition from"information-centered" discourse content to "hearts-centered" discourse source,this research focuses on social emotions,and proactively uses emotional decision-making mechanisms to provide a new research idea for solving the social governance dilemma with technological empowerment.This research addresses the key technical problem of social-emotional security decision from a multidisciplinary perspective,integrating theories and methods from several disciplines such as emotion theory,cognitive psychology,emotional sociology,and computational communication,combined with deep learning algorithms.The main research results are as follows:(1)To address the existing studies that have not focused on public safety governance research from the perspective of social-emotional security,this research proposes a conceptual system of social-emotional security and a methodology framework of decision-making with emphasizing on research paradigm construction.The key prerequisite in this research is to fully understand the features and evolution patterns of social emotions.First,this research starts from the theories related to social-emotional security,bridges the existing research gap between event evolution and emotion identification,and focuses on the need for an emotional security perspective.This research constructs a research framework on the mechanism of social-emotional security,clarifies the structure of social-emotional evolution in the online discourse field of public safety,which consists of inventory emotion,situational emotion and stress emotion,and proposes the theoretical system of social-emotional security.Furthermore,this research constructs a four-dimensional decision analysis structure of"public individual-social group-network discourse field-real discourse field",which is suitable for measuring and evaluating subjects and carriers related to social-emotional security.Then,four key techniques of"cognition-awareness-perception-prediction" are proposed around the emotional decision-making mechanism.This research shifts the"passively feeling" to "actively using",and provides both theoretical and methodological references for the subsequent research on social-emotional security decision-making.(2)To address the lack of theoretical basis and scalability of discrete emotion feature representation in existing studies,this research focuses on social-emotional cognition in online discourse fields and proposes a discrete emotion identification method.This research firstly constructs a discrete emotion recognition model for the public safety domain based on emotion theory and manual annotation experience.This research proposes a discrete emotional appraisal path adapted to the content of public safety discourse,and clarifies the appraisal conditions of specific emotional dimensions.Secondly,this research proposes a correction strategy based on appraisal theory to reveal the emotion dimensions in the public safety domain,and presents a discrete emotion lexicon construction method that integrates emotion validity to quantify the emotion characteristics.To address the problems of loading redundancy of emotion lexicon of multidimensional emotion valences,this research proposes a sensitivity optimization strategy for compound emotion volatility dynamics,and selects the optimal threshold by using the volatility pattern of emotion evolution.Finally,the effectiveness of the proposed method is demonstrated by verifying the correlation between discrete emotional features and the evolutionary dynamics of real events using mutually independent real data.(3)To address the problems of abnormal volatility and insufficient interpretability of multidimensional emotion time series that have not been solved by existing models,this research targets the phenomenon of transient abrupt changes in emotion generated by situational stimuli and proposes a stress emotion discovery method.Based on the continuous multidimensional discrete emotion time series,this research proposes an emotional valence oriented emotional stress time point discovery model.Firstly,by using the LSTM-based attention mechanism encoder-decoder model,the "normal" emotion time series is reconstructed,and the optimal threshold of the stress level is constructed by the maximum likelihood estimation algorithm to accurately estimate the point of abrupt change in emotion.The accuracy and recall of the model prediction are used as the balance point for capturing the emotional dimension of stress and locating the abnormal score.Then,a stress trigger awareness tracing technique is proposed to construct the trigger candidate mechanism to perceive the emotional abnormalities by retracing the candidate events in the window of the stress time interval,and sequentially comparing the similarity between the event emotional representation and the overall social-emotional representation.This method demonstrates the feasibility of tracing the triggering stress emotional risk events with an emotional perspective.(4)To address the lack of evaluation metrics and predictability in existing studies that have not yet measured social-emotional security situations,this research looks at the impact of social emotions on public safety situations and proposes a social-emotional security situation assessment method.This research constructs an index system for social-emotional security situation assessment,extracts three evaluation metrics with seventeen key attributes that can be evaluated and quantified.This research proposes a social-emotional security situation assessment technique,which reflects the social-emotional security index.First,based on the improved Transformer model,the prediction of multidimensional emotional distribution is realized,and the emotional bias metric is evaluated with the proposed method of multidimensional emotional volatility entropy;the situational risk metric is evaluated based on the public’s emotional communication behavior;and the potential hazard metric is evaluated by determining the content security similarity between the target event and the historical events.Then,the three metrics are jointly used to achieve the goal of accurately calculating the social-emotional security index.The experiment results not only prove that the proposed algorithm can achieve the situational prediction for similar events in the public safety field,but also has good reliability in predicting other public safety domains for an additional 30 independent events.Finally,based on the analysis of the social-emotional security index in the above experimental results,a five-level emotion regulation strategy is proposed to justify the social-emotional regulation assessment task.(5)To address the lack of emotional decision-making tools to achieve the lack of enforceability of governance enhancement,this research focuses on the goal of contributing to the overall emotional security and proposes a governance effectiveness evaluation method for social-emotional security.Based on the multiple governance subjects in the emergent situation,a situational decision-making method that contributes to the social-emotional security index is proposed.In terms of governance tools,the social-emotional security causal relationship is constructed,which both eliminates the adverse effects of different subject exposures in multiple measures on social emotions and uses effective exposures to achieve the boosting effect on the social-emotional security index.The experimental results demonstrate that the integration of cause-effect inference techniques enhances the accuracy of social-emotional security situation research and provides the best regulation conditions applicable to enhance the social-emotional security index at the governance content level.Based on this,this research proposes an governance strategy to construct a cyberspace emotional situation defense mechanism.This research also proposes an guidance effect evaluation function and a global effect evaluation functions to facilitate relevant management departments to monitor the effectiveness of governance.The results demonstrate the reliability of the proposed method for developing emotional governance and decision-making solutions.In summary,this research realizes the technical implementation related to focused multidimensional discrete emotional decision-making from the perspective of social-emotional security,which helps to enrich and lead the research direction in the field of public safety and provides decision support for the governance of social-emotional security online discourse fields. |
