一种用于问答系统的加权目标图方法  

A Weighted Target Graph Method for Question Answering System

作　　者：唐卫玲 胡广朋[1] TANG Weiling;HU Guangpeng(School of Computer Science and Engineering,Jiangsu University of Science and Technology,Zhenjiang 212003)

机构地区：[1]江苏科技大学计算机学院,镇江212003

出　　处：《计算机与数字工程》2022年第4期817-820,共4页Computer & Digital Engineering

摘　　要：问答系统是根据用户的提问,从知识库中识别出相应的问题,然后将其答案返回给用户的一种高级检索方式。为了克服输入问句不完整、语义表述不清等对问答系统的影响,论文提出了一种加权目标图的规划识别方法,该方法通过对输入问句的关键词的识别分析实现系统的问题匹配。实验表明,这种基于加权目标图的问题匹配方法可在一定程度上提高匹配的准确率,进而达到提升问答系统性能的作用。The question answering system is an advanced retrieval method that identifies corresponding questions from the knowledge base based on the user’s questions and then returns the answers to the user.In order to overcome the impact of incomplete input questions and unclear semantic expressions on the question and answer system,this paper presents a method for plan recognition of weighted target graphs.This method realizes the system’s problem matching by identifying and analyzing the keywords of the input question sentence.Experiments show that this method of question matching based on weighted target graph can improve the accuracy of matching to a certain extent,and then improve the performance of question answering system.

关 键 词：规划识别 问答系统 加权目标图 词语权重 相似度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]