Preliminary Conference Agenda

摘要：[目的/意义]命名实体识别是自然语言处理任务中众多下游任务的重要基础步骤。古籍作为中华文明的载体，在众多方面具有独特性，为了提高古籍文本中实体识别的准确率，助力古籍资源的智能利用和开发。[方法/过程]首先，本文选取课题组精加工的《二十四史》作为原始数据集，使用gujibert_fan预训练模型对SEQ、SEQ_CRF、SPANPRED模型进行训练，对古籍文本中的实体进行预测。其次，使用了强化实体识别的方法，将预测的实体分为正确和错误两类，生成预测数据集，旨在提供更多上下文和语境信息，以提高实体识别性能。最后，本文引入了合并和多数投票机制的方法（MU）与预测数据集进行整合并训练，以便更准确地捕捉实体边界和类型。[结果/结论]通过评估指标来验证模型的效果，实验结果表明，合并方法的加入使得实体识别的召回率有显著提升，多数投票机制方法提升了模型的F1值。

Based on socio-technical system theory, this study aims to unravel the interaction effect of technical attributes (offering explainable artificial intelligence, XAI) and social factors (algorithm aversion, trust propensity) on human trust to AI and AI-assisted decision-making performance. This study takes a lab experiments approach in the scenario of cross-border e-commerce product selection task. A cross-border e-commerce product selection experiment platform was designed, and 78 candidates were recruited. Configuration analysis was conducted based on the final 69 valid data sets. The findings indicate that, while offering XAI, the behavioral trust to AI can be improved;no matter the trust propensity being high or low, while they don’t have algorithm aversion perception, the AI-assisted decision-making performance can also be improved.

[Purpose/Significance]Network social organizations is an emerging force that cannot be ignored in the evolution of network public opinion. The research on the problem of cooperative refuting rumors led by government and supplemented by network social organization and the public is of enlightening significance for deepening the mechanism of network cooperative refuting rumors under emergencies.[Methods/Procedures]Based on the evolutionary game theory, this paper constructs a dynamic model of bounded rational network social organizations and the public, analyzes the evolutionary stability strategies and evolutionary paths of both sides of the game under different government supervision, and uses Matlab software combined with novel coronavirus(COVID-19) to simulate the situation [Results/Conclusion]The results show that the initial willingness of the players has an impact on the strategy selection, and appropriate rewards and punishments, reducing the misjudgment rate of information review, strengthening the supervision of online public opinion situation and reducing the extra cost of high-quality rumor information can effectively promote the online social organizations to publish high quality rumor information; Guiding by government policies, cultivating public information literacy, strengthening supervision, distinguishing rewards and punishments, and optimizing the government information audit mechanism are conducive to giving full play to the synergy between online social organizations and the public, and improving the efficiency of government network rumor dispelling under emergencies.

[目的/意义] 为考察数据协同情境中社交媒体知识共享的影响因素及组态特征，缓解数据协同主体间存在的知识不平等障碍，以推动数据协同顺利运行和数据价值有效开发。 [方法/过程]本文融合协同技术采纳模型和技术信任理论，综合采用定量和定性比较分析方法，构建并检验了数据协同情境中社交媒体知识共享影响因素的理论模型。[结果/结论]量化分析表明，价值效果预期、使用成本预期和社群环境影响对知识共享意愿产生影响，知识共享意愿、技术信任及应用便利条件影响知识共享行为。同时，任务技术匹配和技术使用经验均对价值效果预期有正向预测作用，技术使用经验对使用成本有负向预测作用。定性比较分析发现，产生强知识共享行为的原因可归纳为参与者驱动型、技术信任-交互便利推动型和低成本驱动型，共三种主要类型。

用户信息素养在谣言验证过程中扮演着至关重要的角色，对于社交媒体谣言治理和网络空间净化具有重要意义。然而，不同信息素养能力的组合如何协同、动态地促进谣言验证意愿，仍缺乏相关的理论研究。本文采用组态视角，通过进行一项两期的纵向研究调查，旨在探索用户在短期和长期时间框架下实现高谣言验证意愿的前因条件组态，并比较它们之间的差异。通过模糊集定性比较分析（fsQCA）的数据分析，本研究发现在短期时，高谣言验证意愿存在三种不同的条件组态：核心技能型、规范应用型和批判反思型，而在长期时，只存在一种条件组态：用户主导与环境辅助型。本研究为信息素养与谣言验证相关研究提供了新的视角和方法启示，也为信息素养教育与社交媒体谣言治理提供了重要的参考依据。