面向数字人文的人物分布式语义表示研究——基于CBDB数据库和古籍文献  

Distributed Representation Learning for the Historical Figures Based on CBDB and Ancient Books: A Digital Humanistic Perspective

作　　者：潘俊 Pan Jun(Department of Big Data,School of Science,Zhejiang University of Science and Technology)

机构地区：[1]浙江科技学院理学院大数据科学系,浙江杭州310023

出　　处：《图书馆杂志》2020年第8期94-102,共9页Library Journal

基　　金：教育部人文社会科学研究青年基金项目“基于知识库和大规模文本的词汇语义表示研究”(项目编号:18YJCZH137);杭州市哲学社会科学规划课题(项目编号:Z20JC060)的研究成果之一。

摘　　要：开放互联环境下的人文研究亟须海量数据资源和新型研究方法。基于中国历代人物资料库(CBDB)数据库和古籍文献采集数据并构建历史人物关系网络,提出一种改进的融合人物影响力的网络表示学习方法,将历史人物表示成具有语义的低维实向量。基于数字人文理念,对人物相关度计算和人物关系挖掘等人文计算任务展开实证研究。通过将网络表示学习引入历史社会网络分析,可为研究者在海量人文数据中挖掘知识、发现问题提供辅助,对丰富和扩展数字人文视角下的人文社科研究范式和研究思维具有积极意义。Humanities research in open and interconnected environment is in an urgent need of massive data resources and new approaches. This study constructs a historical social network based on a large number of data collected from the CBDB knowledge base and ancient literature, and transforms historical figures into low real-value vectors by network representation learning algorithm, which has been specifically improved for historical network by incorporating the nodes influence. Based on the concept of digital humanities, this paper carries out an empirical study on humanities computing tasks, such as person relatedness calculation and relationship mining. This work is the first attempt to introduce the network representation learning technique into the historical social network analysis, which will help researchers to mine knowledge or find clues from massive humanistic data. It will have a positive influence on the expansion of the paradigm and methodology of humanistic research.

关 键 词：数字人文 分布式表示 网络表示学习 社会网络分析 

分 类 号：G250.7[文化科学—图书馆学]