改进的句子相似度计算在问答系统中的应用  

Improved Method for Sentence Similarity Computing and Application in Question Answering System

作　　者：秦元巧[1] 孙国强[1] 

机构地区：[1]上海理工大学,上海200093

出　　处：《微计算机信息》2011年第8期206-208,共3页Control & Automation

摘　　要：本文设计并实现了一种基于词形词序词长、编辑距离和依存文法相集合的一种句子相似度计算方法。通过将顾客输入的自然语言问句与常问问题库中的候选问题集进行相似度计算,自动返回最匹配的答案给顾客,自动更新和维护常问问题库。实验结果表明该方法在问句匹配上比传统方法具有更高的准确率。This paper designed and implemented an approach for computing sentence similarity based on morphology,word order,word length,improved Edit-Distance and Dependency Grammar.This system automatically computes sentence similarity between users ＇ questions in natural language and questions in Question Answering System,and returns the answer to the user.It can also automatically update and maintain Question Answering System.Test result shows that compared with traditional sentence similarity model,this method has more accuracy in matching questions.

关 键 词：句子相似度 问答系统 改进编辑距离 依存文法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]