融合注意力及句法的短文本相似度计算方法  

Short text similarity calculation method combining attention and syntactic

作　　者：蔚佳璇 张起贵 YU Jia-xuan;ZHANG Qi-gui(College of Information and Computer,Taiyuan University of Technology,Jinzhong 030600,China)

机构地区：[1]太原理工大学信息与计算机学院,山西晋中030600

出　　处：《计算机工程与设计》2022年第11期3127-3134,共8页Computer Engineering and Design

基　　金：山西省基础研究项目自然科学基金项目(2013011017-3);太原理工大学科技创新基金项目(9002-03011843)。

摘　　要：针对短文本相似度计算中无法充分利用文本中不同部分的不同重要性且较少考虑句法信息而导致准确率较低的问题,提出一种融合注意力及句法信息的短文本相似度计算方法。通过在语义相似度部分设计混合注意力结构体系,准确提取交互信息,凸显不同粒度的重要程度;在句法结构模块提出一种基于图的依存分析方法准确分析文本的依存关系,构建关系矩阵,计算结构相似度;对两者进行加权融合。实验结果表明,该方法获得的准确率及F1值有所提高,结果优于其它文本相似度方法。Aiming at the problem that the different importance of different parts of the text cannot be fully utilized in the calculation of short text similarity and the syntactic information is less considered,which results in low accuracy,a short text similarity calculation method that combined attention and syntactic information was proposed.By designing a hybrid attention structure system in the semantic similarity part,interactive information was accurately extracted,and the importance of different granula-rities was highlighted.In the syntactic structure module,a graph-based dependency analysis method was proposed to accurately analyze the dependency relationship of the text,construct the relationship matrix,and calculate the structural similarity.A weighted fusion of the two was carried out.Experimental results show that the accuracy and F1 value obtained using this method are improved,and the results are better than that of other text similarity methods.

关 键 词：短文本相似度 混合注意力机制 依存句法分析 关系矩阵 加权融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]