基于情景相似度的口述历史资源事件自动识别研究

A Research on the Automatic Recognition of Events in Oral History Resources Based on Situational Similarity

口述历史资源是个人记忆与社会记忆结合的一种资源形式,其形成过程也是社会记忆再生产的过程,具有典型的叙事特征。对口述历史资源中的名人历史事件进行语义建模,并在元数据层面开展名人历史事件识别具有重要的理论意义与实践意义。文章在构建语义模型的基础上,优化一种计算口述历史资源事件情景相似度的方法,以李政道和CUSPEA事件为例,构建实例事件的情景单元,提取情景要素后对李政道特藏资源中的口述历史资源进行实验。实验结果表明,在设置阈值为0.6时,判断准确率为76.19%,能较大提高此类资源事件标识的效率。

Oral history resources are a form of resource that combines personal memory with social memory.The formation of these resources is also a process of social memory reproduction,characterized by distinctive narrative features.Semantic modeling of celebrity historical events within these archives,and the identification of such events at the metadata level,are of significant theoretical and practical importance.This paper optimizes a method for calculating situational similarity of event scenarios in oral history resources based on the construction of a semantic model.Taking the cases of Tsung Dao Lee and the CUSPEA event as examples,it constructs situational units for instance events and extracts situational elements for experiments with the oral history resources in Tsung Dao Lee's special collection.The experimental results show that with a threshold set at 0.6,the accuracy rate is 76.19%,significantly enhancing the efficiency of event identification in such resources.