基于自然语言的中文地址匹配研究  

Research on Chinese address matching based on natural language

作　　者：徐兵 石少青 陈超 XU Bing;SHI Shao-qing;CHEN Chao(China Southern Power Grid Co.,Ltd.,Guangzhou 510623,China)

机构地区：[1]中国南方电网有限责任公司,广东广州510623

出　　处：《电子设计工程》2020年第16期7-10,16,共5页Electronic Design Engineering

基　　金：中国南方电网科技项目(ZBKJXM20170232)。

摘　　要：针对当前在电力中文地址匹配中存在部分地址歧义的问题,结合自然语言处理的基本原理,提出一种基于贝叶斯算法的中文地址精确匹配方法。为实现该方法,采用贝叶斯算法对中文字符串要素进行切分,以提高地址匹配的精度;采用句法分析,完成对中文地址的层级排序;采用语义标注,完成中文地址的匹配。最后,以上海电力局的部分中文地址数据为样本,采用上述方法进行划分,得到完整和准确的中文地址,并且在不同规模语料库下,本方法得到的分类正确率都在80%以上,具有较高的匹配精度。Aiming at the problem of partial address ambiguity in power Chinese address matching,combining with the basic principle of natural language processing,an accurate Chinese address matching method based on Bayesian algorithm is proposed.In order to achieve this method,Bayesian algorithm is used to segment Chinese string elements to improve the accuracy of address matching;syntactic analysis is used to complete the hierarchical ordering of Chinese addresses;semantic annotation is used to complete the matching of Chinese addresses.Finally,taking some Chinese address data of Shanghai Electric Power Bureau as samples,we use the above method to get complete and accurate Chinese address.In different scale corpus,the classification accuracy of this method is more than 80%,which has high matching accuracy.

关 键 词：自然语言 中文地址 电力营销系统 语义标注 

分 类 号：TN99[电子电信—信号与信息处理]