PSYCH OpenIR  > 社会与工程心理学研究室
基于历史语料库的明君昏君形象定量探析
其他题名A Quantitative Studv of the Images of Enlightened Emperors and Fatuous Emperors Based on Historical Corpus
萧焕仪
导师朱廷劭
2022-06
摘要在我国两千多年浩瀚的历史中,封建帝王们一直被人们津津乐道。人们对哪些皇帝是明君,哪些皇帝是昏君,众说纷纭。有的帝王的形象在人们心中根深蒂固,而有的却不断地擅变。至今为止,哪些皇帝的明君、昏君形象得到大范围的认可,深入民心?他们形象形成的原因是什么?本研究基于历史语料库正史中记载的皇帝语言,从心理语义和人格特征的角度,用数据量化地勾勒明君、昏君皇帝群像图,准确把握这两组皇帝的群体特征的差异,深入认识明君、昏君形象形成的原因,力图突破现有皇帝形象研究的局限。 在心理语义特征方面,本研究首先根据应收尽收的原则,收集市面上明确评价帝王的42本书籍,对帝王按评价其书籍的数量进行了排名,得到了“十大明君名单”和“十大昏君名单”。然后收集正史中这些具有明君、昏君形象的皇帝们的口头言语和诏书,运行CC-LIWC软件,定量地计算出这两组皇帝使用具有特定心理语义的词类的频率,并结合心理学、历史学,探析它们的差异。结果表明,中国明君使用“第三人称(单数和复数)代词”显著高于昏君,而且使用“健康”的词类的频率边缘显著高于昏君。而昏君使用“时态标志”和“关注过去”词类的频率边缘显著高于明君。 在大五人格特征方面,本研究率先构建了基于RKT相关知识迁移的中国古代帝王大五人格预测模型。该模型首先将相关知识从知识空间S迁移到文本空间,然后构建特征空间。另一方面,通过相关知识,构建虚拟标签空间。最后,通过组合特征空间和标签空间构建虚拟训练数据集,供给DANK学习。它利用了很多相关知识实现迁移学习,包括历史知识一中国帝王自称“联”,统计学相关系数的知识,以及心理学知识一人格分数服从正态分布等,来构造虚拟训练数据集,大幅度减少了人工标注的工作量。RKT模型还因为具备了域对抗神经网络(DANK)的良好学习能力,提升了人格预测的准确度,平均RM SE从9.16降至5.40,可作为迁移学习的实践案例供研究者参考。本RKT模型可以进一步应用于预测中国古代特定群体或时期的古人的大五人格,对未来的特定群体的人格研究开辟了新的途径。 本研究应用RKT模型预测元、明、清三代帝王的大五人格分数,并结合史实对三朝帝王群像进行梳理,定量和定性相结合地比较分析三代帝王群体人格分数差异的原因,进一步确认RKT模型的准确性。本研究率先对帝王群像采用定量的比较研究,有助于突破历史学定性分析的局限性,为心理历史学研究的开展提供了新颖的思路与有益的借鉴。 本研究应用RKT模型定量预测具有明君、昏君形象的两组帝王的大五人格,并进行差异的显著性检验。结果显示,明君、昏君人格五个维度的分数无显著差异,说明他们的形象形成不仅是他们自身人格这单一因素所致,而是还有更多内因、外因等综合因素,因时因势,时代造就。本研究有助于人们从新的角度认识明君、昏君,加强对他们形象形成原因的理解,也为未来的研究者从更长时间段内探讨中国人的心理特征及变化趋势提供了新的视角。
其他摘要In Chinese vast history of more than 2000 years, feudal emperors have always been talked about by people. People have different opinions on which emperors are enlightened emperors and which emperors are fatuous emperors. Some emperors' images are deeply rooted in people's hearts, while others are constantly evolving. So far, which emperors' images of enlightened emperors and/or fatuous emperors have been widely recognized and deeply rooted in the hearts of the people? What are the reasons for the formation of their images? Based on the emperors' language recorded in the official history of the historical corpus, from the perspective of psychological semantics and personality traits, this study uses data to quantitatively outline the group images of enlightened emperors and fatuous emperors, accurately grasp the differences of the group characteristics of the two groups of emperors, deeply understand the reasons for the formation of the images of enlightened emperors and fatuous emperors, and try to break through the limitations of the existing research on the image of emperor. In terms of psychological semantic features, firstly, according to the principle of "collecting all accounts receivable", we collected 42 books that clearly evaluated the emperors in the market, ranked the emperors according to the number of books evaluated, and obtained the lists of "the top ten enlightened emperors" and "the top ten fatuous emperors". Secondly, we collected the spoken utterances and imperial edict records of these two groups of emperors from the official history books. Thirdly, we applied the Classical Chinese Linguistic Inquiry and Word Count(CC-LIWC)software, quantitatively calculate the frequency of the two groups of emperors using word categories with specific psychological semantics, and analyze their differences in a psychological and historical perspective. The results show that enlightened emperors used "third person(singular and plural)pronouns" significantly more than fatuous emperors, and used "health" word categories marginally significantly more than fatuous emperors. While the use of "tense marker" and "past focus" by the fatuous emperors is marginally significantly more than that of the enlightened emperors. In terms of Big Five personality traits, this paper takes the lead in constructing a prediction model of Big Five personality of ancient Chinese emperors based on relational knowledge transfer(RKT)learning. The model first migrates relational knowledge from knowledge space s to text space, and then constructs feature space. On the other hand, the virtual label space is constructed through relational knowledge. Finally, the virtual training data set is constructed by combining feature space and label space for DANN learning. It used a lot of relevant knowledge to implement transfer learning, including Chinese emperors' self-claimed character "then", the knowledge of statistical correlation coefficient, and psychological knowledge including personality scores having normal distribution, to produce the virtual training datasets, which greatly reduces the workload of manual labeling. RKT model also equips with the good learning ability of Domain-Adversarial Training of Neural Network(DANN), which improves the accuracy of personality prediction, and the average RMSE decreases from 9.16 to 5.40. It can be used as a practical case of transfer learning for researchers' reference. This RKT model can be further applied to predict the Big Five personality of ancient Chinese in a specific group or period. This RKT research method opens up a new way for the personality research of specific groups in the future. The RKT model was used to predict the Big Five personality scores of the emperors of the Yuan, Ming and Qing Dynasties in this study. Combined with historical facts, this paper combed through the group images of the emperors of the three dynasties, compared and analyzed the reasons for the differences in the personality scores of the three generations of emperors quantitatively and qualitatively, and further confirmed the accuracy of the RKT model. This study takes the lead in using quantitative comparative research on the group images of emperors, which helps to break through the limitations of historical qualitative analysis and provides novel ideas and useful reference for the research of psychological history. In this study, RKT model was used to quantitatively predict the Big Five personalities of the two groups of emperors with the images of enlightened emperors and fatuous emperors, and the significance of the difference is tested. The results showed that there was no significant difference in the scores of the five dimensions of enlightened emperors and fatuous emperors' personalities. It means that their image formation is not only caused by the single factor of their own personalities, but also by more comprehensive factors such as internal and external factors, which are formed during the times. This study will help people understand the enlightened emperors and fatuous emperors from a new perspective and strengthen the understanding of the reasons for their images formation. It provides a new perspective for exploring the psychological characteristics and change trend of Chinese people in a longer period of time for the future researchers.
关键词迁移学习 昏君 大五人格 CC-LIWC 元明清朝
学位类型硕士
语种中文
学位名称理学硕士
学位专业应用心理学
学位授予单位中国科学院大学
学位授予地点中国科学院心理研究所
文献类型学位论文
条目标识符https://ir.psych.ac.cn/handle/311026/45009
专题社会与工程心理学研究室
推荐引用方式
GB/T 7714
萧焕仪. 基于历史语料库的明君昏君形象定量探析[D]. 中国科学院心理研究所. 中国科学院大学,2022.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
萧焕仪-硕士学位论文.pdf(1891KB)学位论文 开放获取CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[萧焕仪]的文章
百度学术
百度学术中相似的文章
[萧焕仪]的文章
必应学术
必应学术中相似的文章
[萧焕仪]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。