改进Jaro-Winkler算法在迎宾机器人语音交互中的应用  

Application of Improved Jaro-Winkler Distance in Speech Interaction of Reception Robot

作　　者：吴凌芬 杨小渊[2] 叶添杰[2] 刘冰[2] 王太宏[2] 

机构地区：[1]厦门大学信息科学与技术学院,厦门361005 [2]厦门大学萨本栋微米纳米科学技术研究院,厦门361005

出　　处：《现代计算机（中旬刊）》2015年第3期8-13,共6页Modern Computer

基　　金：国家自然科学基金(No.61376073)

摘　　要：针对Jaro-Winkler算法在计算两个字符串的相似度时只考虑字符的换位数目,未考虑字符插入和删除编辑操作的问题,提出一种基于Levenshtein算法改进Jaro-Winkler算法的方法。通过改进相似度的计算公式和调整关联系数pd,实现Jaro-Winkler算法的改进,提高字符串匹配的准确度。基于内容丰富的本地文本数据库,将改进Jaro-Winkler算法应用于迎宾机器人语音交互中的字符串匹配,其正确率大于96%。测试结果表明,迎宾机器人能够语音交互更多的内容和快速地应答提问,并实现特定的迎宾动作。When measuring the similarity of two strings, Jaro-Winkler distance only considers the number of transpositions, without considering the insertions and deletions. Aiming at this problem, proposes an improved Jaro-Winkler distance method based on the Levenshtein distance.Modifies the formula of similarity and adjusting the correlation coefficient of pd to improve Jaro-Winkler distance, and the accuracy of the strings matching is increased. Based on the local database with a large number of text strings, applies the improved Jaro-Winkler distance to achieve the strings matching in speech interaction of reception robot, and its accuracy is higher than 96%. Test results show that reception robot can realize more words and quickly reply to questions by speech interaction with specific welcome actions.

关 键 词：Jaro-Winkler算法 迎宾机器人 语音交互 字符串匹配 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]