融合多特征的中文产品命名实体识别  

Fusion of Multiple Features for Chinese Product Named Entity Recognition

作　　者：谷川[1] 周宏宇[1] 于江德[1] 

机构地区：[1]安阳师范学院计算机与信息工程学院,安阳455000

出　　处：《科学技术与工程》2013年第31期9417-9421,9435,共6页Science Technology and Engineering

基　　金：国家自然科学基金(60875081);河南省基础与前沿技术研究计划项目(112300410182);河南省软科学研究项目(132400410248)资助

摘　　要：针对中文产品命名实体,提出了一种基于多种特征融合的识别方法。以词为标注粒度,将多种特征融合到条件随机场模型中,采用递增式学习策略选取最优的特征模板,实现了从中文自由文本中识别产品命名实体。实验表明,该方法获得了令人满意的实验效果,准确率、召回率和F值分别达到94.87%、92.50%和93.67%。A approach for Chinese product named entity recognition based on fusion of multiple features is pres- ented. The approach fuses multiple features to conditional random fields model, using Chinese word as processing unit, which adopts incremental learning strategy to optimize feature templates and recognizes product named entity from Chinese free text. The experimental results show that the approach performs quite well, the precision, recall and F value can reach 96. 87%, 94. 50% and 95. 67% respectively.

关 键 词：多特征 产品命名实体识别 条件随机场 

分 类 号：TP391.11[自动化与计算机技术—计算机应用技术]