自然语言语义相关度计算模型的k枝剪求解法  

K-pruning algorithm for semantic relevancy calculating model of natural language

作　　者：刘运通[1] 梁燕军[1] 

机构地区：[1]安阳师范学院计算机与信息工程学院,河南安阳455000

出　　处：《计算机工程与设计》2013年第8期2939-2943,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(U1204402);河南省教育厅科学技术研究重点基金项目(13B520894)

摘　　要：为了能够更为合理地利用语义来进行自然语言处理,提出了一种自然语言语义相关度计算模型及该模型的k枝剪求解法。在该模型中使用语句的语义相关度来判定最佳语法分析方案;分析了语句的两层语义结构并给出了其数学描述方法;在模型求解过程中,会形成一个状态空间树,使用k枝剪法舍弃可能性较小的状态,可以有效地降低计算复杂度并较为准确地计算出模型的近似解。实验结果表明,该方法具有一定的可行性。In order to use semantics more effectively in natural language processing,a semantic relevancy calculating model of natural language is proposed,and the k-pruning algorithm is proposed for solving the model.In the model,the best parsing process for a clause could be determined by the value of semantic relevancy of the clause.The two-level semantic structure of a clause are analyzed,and two grammar rules are used to describe the two-level semantic structure.In the process of solving the model,a state tree would be generated;the k-pruning algorithm could be used to delete the states with less semantic relevancy when searching the state tree,and the computational complexity is effectively reduced and the approximate solution could be accurately acquired.In the experiments,the results demonstrate that the algorithm is effective in solving the model.

关 键 词：自然语言 语义相关度 状态空间树 k枝剪法 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]