摘要
新兴主题突变前兆特征识别,对于预测发展方向及制定技术战略具有重要意义,它能够助力相关领域提前把握发展动态,在竞争中占据先机,增强整体竞争力。首先,以科技论文发表时间划分时间切片,基于BERT-Topic和Louvain算法分别生成不同时期微观、宏观维度的主题信息;其次,依据相似度实现双维度主题关系演变;然后,绘制双维动态知识网络并定义新兴主题突变特征,依据突变特征构建主题新生涌现指标和主题交叉演变指标分析结构突变;最后,利用网络结构熵、异常信号等方法进行突变前兆特征识别。以干细胞领域为例进行实证研究,发现了主题突变前网络结构会出现显著波动或异常的前兆特征,通过理论依据和已有成果验证了新兴主题突变前兆特征识别方法的有效性和可行性,为提前捕捉新兴主题突变提供新视角和新方法。
Identifying early signs of emerging thematic changes is of great significance for predicting future trends and formulating technical strategies.It can help relevant fields grasp development dynamics in advance,gain a competitive edge,and enhance overall competitiveness.First,time slices are defined based on the publication dates of scientific papers.Using BERT-Topic and the Louvain algorithm,thematic information at both micro and macro levels is generated for different time periods.Second,the evolution of thematic relationships across both dimensions is analyzed based on similarity.Third,a two-dimensional dynamic knowledge network is constructed,and emerging thematic mutation characteristics are defined.Based on these characteristics,indicators for the emergence of new themes and the cross-evolution of themes are developed to analyze structural mutations.Finally,the network structural entropy,anomalous signals,and other methods are used to identify mutation precursor features.Using stem cell research as an example,empirical studies have identified precursor characteristics of significant fluctuations or abnormalities in network structures prior to the emergence of new themes.These findings have been validated through theoretical frameworks and existing research outcomes,demonstrating the effectiveness and feasibility of new methods for identifying precursor characteristics of emerging themes.This provides a new perspective and approach for anticipating and capturing emerging theme mutations in advance.
作者
孙宇
王超
许海云
Sun Yu;Wang Chao;Xu Haiyun(Qilu University of Technology(Shandong Academy of Sciences),Jinan 250014;Shandong University of Technology,Zibo 255049)
出处
《情报杂志》
北大核心
2026年第3期139-148,共10页
Journal of Intelligence
基金
国家社会科学基金项目“高潜能未来产业的跨领域技术交叉融合机理及路径研究”(编号:24BTQ072)研究成果。
关键词
突变前兆
新兴主题
主题突变
主题识别
前兆特征
知识网络
双维度
属性结构理论
复杂网络理论
precursors of mutation
emerging themes
thematic mutation
theme identification
precursor characteristics
knowledge networks
two dimensions
attribute structure theory
complex network theory
