基于知识图谱的中文关键短语提取算法  

CHINESE KEY PHRASE EXTRACTION ALGORITHM BASED ON KNOWLEDGE GRAPH

作　　者：徐远威 李劲华 Xu Yuanwei;Li Jinhua(School of Data Science and Software Engineering,Qingdao University,Qingdao 266071,Shandong,China)

机构地区：[1]青岛大学数据科学与软件工程学院,山东青岛266071

出　　处：《计算机应用与软件》2023年第6期250-256,265,共8页Computer Applications and Software

基　　金：山东省重大科技创新工程项目(2019JZZY020101)。

摘　　要：针对传统的中文关键短语算法没有充分考虑到语义关系、涵盖信息量少和准确率低等问题,提出一种基于知识图谱来提取关键短语的算法。该算法运用知识图谱的语义网络结构来挖掘文本中两个词之间的潜在关系,通过赋予边权值具体量化潜在关系,构建集成近邻传播(Affinity Propagation,AP)聚类和图的中心性算法来提取关键短语的框架。实验表明,与多种传统关键短语提取算法相比,该算法在准确率、召回率和F1值上有较明显的提升。The traditional Chinese key phrase algorithms don t fully consider the semantic relationship,less information and low precision.To solve these problems,this paper proposes an algorithm to extract key phrase based on knowledge graph.The algorithm used the semantic network structure of knowledge graph to mine the potential relationship between two words in the text,quantified the potential relationship by giving edge weights,and constructed a framework integrating affinity propagation(AP)clustering and graph centrality algorithm to extract key phrases.Experimental results show that compared with many traditional key phrase extraction algorithms,this algorithm has a significant improvement in precision,recall and F1 value.

关 键 词：关键短语提取 知识图谱 近邻传播聚类 中心性算法 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]