基于兴趣点简称的检索方法研究  

Study of Chinese Points of Interest Search Based on Name Abbreviations

作　　者：毛姝洁[1] 张雪虎[1] 

机构地区：[1]北京大学遥感与地理信息研究所空间智能计算实验室,北京100871

出　　处：《太原理工大学学报》2008年第S1期52-55,共4页Journal of Taiyuan University of Technology

基　　金：国家重点研究基础发展规划(973)项目(2006CB70130000)

摘　　要：提出了一种基于统计方法和分词的匹配模型,该模型能够根据简称找出最可能的全称。这一模型由三个部分组成:针对不同类别的全称,对兴趣点全称进行分类,将全称划分成机构、公司、行业、区域和未知五类单词;把单词按照长度分为六类,然后在此基础上建立隐马尔可夫统计模型;在越频繁使用的全称越有可能产生简称的假设下,建立一个流行度统计模型。实验表明上述模型的首选准确率达到近85%,比现有搜索引擎的匹配能力有了显著的提高。This paper provides a model based on statistical methods and segmentation,which can retrieve the full name matching the abbreviation mostly.This model consists of three parts: The full names of POI were firstly segmented and the segments were classified as one of organizations,companies,industry,location or unknown;the segments were classified into 6 classes according to their length,then,hidden Markov model was built with respect to every class mentioned above;the more popular a POI name is,the more likely...

关 键 词：简称 兴趣点名称匹配 中文信息处理 隐马尔科夫模型 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]