Chinese named entity recognition based on hierarchical hybrid model

Authors:
Zhihua Liao;Zili Zhang;Yang Liu
Affiliations:
Hunan Normal University, Changsha, Hunan, China;Southwest University, Chongqing, China and Deakin University, Geelong, VIC, Australia;Jilin University, Changchun, Jilin, China
Venue:
PRICAI'10 Proceedings of the 11th Pacific Rim international conference on Trends in artificial intelligence
Year:
2010

Citing 5
Cited 1

Chinese named entity identification using class-based language model

COLING '02 Proceedings of the 19th international conference on Computational linguistics - Volume 1
Markov logic networks

Machine Learning
Chinese Named Entity Recognition combining a statistical model with human knowledge

MultiNER '03 Proceedings of the ACL 2003 workshop on Multilingual and mixed-language named entity recognition - Volume 15
Chinese named entity recognition with cascaded hybrid model

NAACL-Short '07 Human Language Technologies 2007: The Conference of the North American Chapter of the Association for Computational Linguistics; Companion Volume, Short Papers
Discriminative training of Markov logic networks

AAAI'05 Proceedings of the 20th national conference on Artificial intelligence - Volume 2

Automatic dominant character identification in fables based on verb analysis - Empirical study on the impact of anaphora resolution

Knowledge-Based Systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

Chinese named entity recognition is a challenging, difficult, yet important task in natural language processing. This paper presents a novel approach based on a hierarchical hybrid model to recognize Chinese named entities. Three mutually dependent stages- boosting, Markov Logic Networks (MLNs) based recognition, and abbreviation detection - are integrated in the model. AdaBoost algorithm is utilized for fast recognition of simple named entities first. More complex named entities are then piped into MLNs for accurate recognition. In particular, the left boundary recognition of named entities is considered. Lastly, special care is taken for classifying the abbreviated named entities by using the global context information in the same document. Experiments were conducted on People's Daily corpus. The results show that our approach can improve the performance significantly with precision of 94.38%, recall of 93.89%, and Fβ=1 value of 93.97%.