Research on Chinese named entity recognition using combined boundary-PoS feature

被引：0

作者：

Qiang, Bao-Hua ^{[1
]}

Huang, Jun ^{[1
]}

Wang, Yu-Feng ^{[2
]}

Wang, Sai ^{[1
]}

Wang, Yong ^{[1
]}

机构：

[1] Gulin Univ Elect Technol, Coll Comp Sci Engn, Guilin, Guangxi, Peoples R China

[2] Technol Grp Corp, Res Inst China Elect 54, Shijiazhuang 050000, Peoples R China

来源：

DESIGN, MANUFACTURING AND MECHATRONICS (ICDMM 2015) | 2016年

关键词：

Named entity recognition; conditional random field; Part of Speech (PoS); word boundary; boundary-PoS feature; feature templates;

D O I：

暂无

中图分类号：

T [工业技术];

学科分类号：

08 ;

摘要：

Named entity recognition (NER) on complex organizations (ORG), locations (LOC) and person names (PER) is a difficult issue in natural language processing. We propose a novel 6 tags method which combines boundary and Part of Speech (6-tag-boundary-PoS) to improve the recognition rate for NER based on conditional random field. Large-scale open tests on real corpus indicate that the 6-tag-boundary-PoS feature can effectively shorten the training time, and the F1 value, precision (P), recall (R) of ORG and LOC are 98.25%, 97.82%, 98.68% and 91.81%, 89.76%, 93.97%, respectively.

引用

页码：839 / 848

页数：10

共 9 条

[1]

Feng Yuan-yong, 2008, Acta Electronica Sinica, V36, P1833

[2]

[冯元勇 FENG Yuanyong], 2008, [中文信息学报, Journal of Chinese Information Processing], V22, P104

[3]

Lafferty John, 2001, INT C MACH LEARN ICM

[4]

[李丽双 Li Lishuang], 2006, [中文信息学报, Journal of Chinese Information Processing], V20, P51

[5]

[李中国 Li Zhongguo], 2006, [中文信息学报, Journal of Chinese Information Processing], V20, P44

[6]

Shi Shuicai, 2013, Computer Engineering and Applications, V49, P147, DOI 10.3778/j.issn.1002-8331.1109-0592

[7]

Yang Xiao-dong, 2011, Computer Engineering, V37, P169, DOI 10.3969/j.issn.1000-3428.2011.08.058

[8]

ZHANG Zhu-yu, 2008, 4 NAT INF RETR CONT, P111

[9]

Zhou Jun-sheng, 2006, Acta Electronica Sinica, V34, P804

← 1 →