基于多算法多因素融合的关键词提取方法  

Chinese Text Keyword Extraction Method Based on Multi-Factor Fusion

作　　者：柴新茹 余宏杰 CHAI Xinru;YU Hongjie(College of Mechanics,Anhui Science and Technology University,Fengyang 233100,China;College of Information Networks,Anhui Science and Technology University,Bengbu 233000,China)

机构地区：[1]安徽科技学院机械工程学院,安徽凤阳233100 [2]安徽科技学院信息与网络工程学院,安徽蚌埠233000

出　　处：《枣庄学院学报》2023年第2期55-61,77,共8页Journal of Zaozhuang University

基　　金：基于区块链技术的乳制品质量安全监测关键技术的研究与应用(202204c06020065);数字农业知识图谱与数字建模探索研究(880640)。

摘　　要：为进一步提升关键词提取准确率,提出基于多算法多特征融合的中文文本关键词提取方法。对现有的TF-IDF算法和TextRank算法进行融合,同时融合词频、词长、词语位置、词性这四种影响因素进行加权。通过试验获取权重公式的相对最优权重系数,对改进后两种算法分别提取出权重值由高到低的前K个候选关键词,最终采取投票法筛选出结果。从准确率P、召回率R、准确率和召回率的加权调和平均值F值三个指标对文中融合改进算法(Fusion-T算法)、经典的TF-IDF算法和TextRank算法进行对比。试验结果表明:算法改进后,P、R、F值分别提高了:6.18%、4.97%、5.99%。In order to further improve the accuracy of keyword extraction,a Chinese-text keyword extraction method based on multi-algorithm and multi-feature fusion is proposed.The existing TF-IDF algorithm and TextRank algorithm are fused,and the four influencing factors of word frequency,word length,word position,and part-of-speech are combined to weight.The relative optimal weight coefficient of the weight formula was obtained through experiments,and the top K candidate keywords with weight values from high to low were extracted for the improved two algorithms,and finally the results were filtered by voting.The three indexes of P(accuracy),R(recall)and F value(weighted harmonic average of P and R)were compared.Experimental results show that after the algorithm is improved,the values of P,R and F increase by 6.18%,4.97%and 5.99%,respectively.

关 键 词：中文信息处理 关键词提取 TF-IDF TextRank 位置加权 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]