Applying conditional random fields to chinese shallow parsing

被引：0

作者：

Tan, YM ^{[1
]}

Yao, TS ^{[1
]}

Chen, Q ^{[1
]}

Zhu, JB ^{[1
]}

机构：

[1] Northeastern Univ, Nat Language Proc Lab, Shenyang 110004, Peoples R China

来源：

COMPUTATIONAL LINGUISTICS AND INTELLIGENT TEXT PROCESSING | 2005年 / 3406卷

关键词：

D O I：

暂无

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

Chinese shallow parsing is a difficult, important and widely-studied sequence modeling problem. CRFs are new discriminative sequential models which may incorporate many rich features. This paper shows how conditional random fields (CRFs) can be efficiently applied to Chinese shallow parsing. We employ using CRFs and HMMs on a same data set. Our results confirm that CRFs improve the performance upon HMMs. Our approach yields the F1 score of 90.38% in Chinese shallow parsing with the UPenn Chinese Treebank. CRFs have shown to perform well for Chinese shallow parsing due to their ability to capture arbitrary, overlapping features of the input in a Markov model.

引用

页码：167 / 176

页数：10

共 20 条

[1]

Abney S. P., 1991, Principle-based parsing, P257, DOI DOI 10.1007/978-94-011-3474-3_10

[2]

[Anonymous], 2002, THESIS U EDINBURGH

[3]

[Anonymous], 1991, THESIS U PARIS 11

[4]

John LaffertyAndrew McCallum Fernando CN Pereira., 2001, CONDITIONAL RANDOM F, DOI 10.5555/645530.655813

[5]

LI SJ, 2003, CHINESE J COMPUTERS, V26, P1734

[6]

MCCALLUM A, 2003, CHINESE WORD SEGMENT

[7]

Mccallum A., 2003, EARLY RESULTS NAMED

[8]

McCallum A., 2000, Icml, V17, P591

[9]

McCallum A.K., 2002, MALLET: A Machine Learning for Language Toolkit

[10]

PAZ A, 1971, ITNRO PROBABILISTIC

← 1 2 →