人名识别技术在中国招中标领域的应用  

Application of name recognition technology to field of bidding in China

作　　者：张悦 潘淑文 刘秀磊[1,3] ZHANG Yue PAN Shuwen LIU Xiulei(Beijing Key Laboratory of Interuet Culture and Digital Dissemination Research, Beijing Information Science ＆ Technology University, Beijing 100192, China Computer Center, Beijing Information Science ＆ Technology University, Beijing 100192, China State key Laboratory of Networking and Switching Technology, B eijing University of Posts and Telecommunications, Beijing 100876,China)

机构地区：[1]北京信息科技大学网络文化与数字传播北京市重点实验室,北京100192 [2]北京信息科技大学计算中心,北京100192 [3]北京邮电大学网络与交换技术国家重点实验室,北京100876

出　　处：《北京信息科技大学学报（自然科学版）》2017年第5期72-76,83,共6页Journal of Beijing Information Science and Technology University

基　　金：国家重点研发计划资助(2016YFC0801407);国家自然科学基金资助(61601039);北京信息科技大学科研基金资助(1625008);网络文化与数字传播北京市重点实验室开放课题资助(ICDD201607);网络与交换技术国家重点实验室(北京邮电大学)开放课题资助(SKLNST-2016-2-08)

摘　　要：招中标数据中的联系人和评审专家是企业关注的关键信息之一,为建立企业竞争信息服务平台,实现招中标数据的自动处理,采用基于隐马尔可夫模型的人名识别算法,并使用招中标数据进行模型修正的方法提高对联系人和评审专家提取的准确率。修正后的人名识别模型准确率从80%提高到了89.5%,召回率从88.9%提高到95.1%,达到了实际中可以应用的准确率要求,具有良好的应用价值。In the bidding area,the person to contact and review experts is one of the key information of concern to the enterprises. To set up Enterprise Competitive Information Service Platform and realize the automatic processing of bidding data,this article applies name recognition method based on hidden Markov model,and uses the bidding data to carry out the model parameters to increase the extraction accuracy of contact person and review experts. After the optimization,the accuracy rate of name recognition is improved from 80% to 89. 5%,and the recall rate is improved from 88. 9% to95. 1%,meeting the actual application requirement of the accuracy rate,with good application value.

关 键 词：人名识别 隐马尔可夫模型 维特比算法 招中标数据分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]