基于异构数据联合训练的中文分词法

doi:10.16180/j.cnki.issn1007-7820.2019.04.007

电子科技 ›› 2019, Vol. 32 ›› Issue (4): 29-33.doi: 10.16180/j.cnki.issn1007-7820.2019.04.007

基于异构数据联合训练的中文分词法

姜猛¹,王子牛²,高建瓴¹

^1. 贵州大学大数据与信息工程学院,贵州贵阳 550025
^2. 贵州大学网络与信息化管理中心,贵州贵阳 550025

收稿日期:2018-03-18 出版日期:2019-04-15 发布日期:2019-03-27
作者简介:姜猛(1994-),男,硕士研究生。研究方向:机器学习、数据挖掘。|王子牛(1961-),男,副教授。研究方向:信息与信号处理、数据挖掘、计算机仿真技术。|高建瓴(1969-),女,副教授。研究方向:云计算、数据库应用。
基金资助:
贵州省科学技术基金(黔科合J字[2015]2045);贵州大学研究生创新基金(研理工2017016)

Chinese Word Segmentation Based on Joint Training of Heterogeneous Data

JIANG Meng¹,WANG Ziniu²,GAO Jianling¹

^1. School of Big Data & Information Engineering,Guizhou University,Guiyang 550025,China;
^2. Network and Information Management Center,Guizhou University,Guiyang 550025,China

Received:2018-03-18 Online:2019-04-15 Published:2019-03-27
Supported by:
Guizhou Science and Technology Fund(Guizhou Science and Technology Agency J [2015]2045);Guizhou University Graduate Innovation Fund(Graduate Science and Technology 2017016)

摘要/Abstract

摘要：

中文分词技术作为中文信息处理中的关键基础技术之一,基于深度学习模型的中文分词法受到广泛关注。然而,深度学习模型需要大规模数据训练才能获得良好的性能,而当前中文分词语料数据相对缺乏且标准不一。文中提出了一种简单有效的异构数据处理方法,对不同语料数据加上两个人工设定的标识符,使用处理过的数据应用于双向长短期记忆网络结合条件随机场(Bi-LSTM-CRF)的中文分词模型的联合训练。实验结果表明,基于异构数据联合训练的Bi-LSTM-CRF模型比单一数据训练的模型具有更好的分词性能。

关键词: 中文分词, 深度学习, Bi-LSTM-CRF, 异构数据, 联合训练, 语料库

Abstract:

Chinese word segmentation technology is one of the key basic technologies in Chinese information processing. The Chinese word segmentation method based on deep learning model is widely concerned. However, the deep learning model requires large-scale data training to obtain good performance, but the current Chinese sub-word data is relatively lacking and the standards are not the same. This paper proposes a simple and effective method of heterogeneous data processing. Firstly, two artificially-set identifiers are added to different corpus data, and then the processed data is applied to the joint training of Bi-LSTM-CRF Chinese word segmentation model. Experimental results show that the Bi-LSTM-CRF model based on heterogeneous data joint training has better segmentation performance than the single data training model.

Key words: Chinese word segmentation, deep learning, Bi-LSTM-CRF, heterogeneous data, joint training, corpus

中图分类号:

TP391

姜猛,王子牛,高建瓴. 基于异构数据联合训练的中文分词法[J]. 电子科技, 2019, 32(4): 29-33.

JIANG Meng,WANG Ziniu,GAO Jianling. Chinese Word Segmentation Based on Joint Training of Heterogeneous Data[J]. Electronic Science and Technology, 2019, 32(4): 29-33.

图/表 4

图1

表1

表2

表3

参考文献 16

[1]	郑捷 . NLP汉语自然语言处理原理与实践[M]. 北京: 电子工业出版社, 2017.
	Zheng Jie. Principles and practice of NLP Chinese natural language processing[M]. Beijing: Publishing House of Electronics Industry, 2017.
[2]	Hinton G E, Salakhutdinov R R . Reducing the dimensionality of data with neural networks[J]. Science, 2006,313(5786):504-507. doi: 10.1126/science.1127647 pmid: 16873662
[3]	Collobert R, Wesotn J, Bottou L . Natural language processing(almost) from scratch[J]. The Journal of Machine Learning Research, 2011(12):2493-2537.
[4]	Zheng X, Chen H, Xu T. Deep learning for Chinese word segmentation and POS tagging [C].Seattle:Conference on Empirical Methods in Natural Language Processing, Association for Computational Linguistics, 2013.
[5]	Chen X, Qiu X, Zhu C. Gated recursive neural network for Chinese word segmentation [C]. Beijing:Proceedings of the 53 ^rd Annual Meeting of the Association for Computational Linguistics and the 7 ^th International Joint Conference on Natural Language Processing,Association for Computational Linguistics , 2015.
[6]	Chen X, Qiu X, Zhu C. Long short-term memory neural networks for Chinese word segmentation[C]. Lisbon:Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing,Association for Computational Linguistics, 2015.
[7]	Yao Yushi, Huang Zheng. Bi-directional LSTM recurrent neural netword for Chinese word segmentation[C]. Kyoto:International Conference on Neural Information Processing,Springer, 2016.
[8]	Jiang Wenbin, Huang Liang, Liu Qun. Automatic adaptation of annotation standards:Chinese word segmentation and POS tagging[C]. Morristown: In the Joint Conference of the 47 ^th Annual Meeting of the ACL and the 4 ^th International Joint Conference,Association for Computational Linguistics , 2009.
[9]	Sun Weiwei, Wan Xiaojun. Reducing approximation and estimation errors for Chinese lexical processing with heterogeneous annotations[C]. Jeju Island:Proceedings of the 50 ^th Annual Meeting of the Association for Computational Linguistics,Association for Computational Linguistics,2012.[10]Qiu Xipeng,Zhao Jiayi,Huang Xuanjing.Joint Chinese word segmentation and POS tagging on heterogeneous annotated corpora with multiple task learning[C].Seattle:Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing,Association for Computational Linguistics , 2013.
[11]	Li Zhenghua, Chao Jiayuan, Zhang Min, et al. Coupled sequence labeling on heterogeneous annotations: POS tagging as a case study[C]. Beijing:In Proceedings of the 53 ^rd Annual Meeting of the Association for Computational Linguistics and the 7 ^th International Joint Conference on Natural Language Processing. Association for Computational Linguistics , 2015.
[12]	Chen Xinchi, Shi Zhan, Qiu Xipeng , et al. Adversarial multi-criteria learning for Chinese word segmentation[C]. Vancouver:Proceedings of the 55 ^th Annual Meeting of the Association for Computational Linguistics,Association for Computational Linguistics , 201l.
[13]	Melvin Johnson, Mike Schuster, Quoc V Le . Google’s multilingual neural machine translation system-enabling zero-shot translation[J]. Computational Linguistics, 2017(5):339-351. doi: 10.1162/tacl_a_00065
[14]	王贵新, 郑孝宗 . 基于Word2vec的短信向量化算法[J]. 电子科技2016, 29(4):49-52. doi: 10.16180/j.cnki.issn1007-7820.2016.04.013
	Wang Guixin, Zheng Xiaozong . An algorithm for vectoring SMS based on Word2vec[J]. Electronic Science and Technology, 2016,29(4):49-52. doi: 10.16180/j.cnki.issn1007-7820.2016.04.013
[15]	Zheng Xiaoqing, Chen Hanyang, Xu Tianyu. Deep learning for Chinese word segmentation and POS tagging[C]. Seattle:Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, Association for Computational Linguistics, 2013.
[16]	Cai Deng, Zhao Hai. Neural word segmentation learning for Chinese[C]. Berlin:Proceedings of the 54 ^th Annual Meeting of the Association for Computational Linguistics , 2016.

语料库	标识符
PKU	<pku>江泽民发表新年贺词</pku>
MSR	<msr>穆守家两间低矮的小屋里</msr>
AS	<as>民族所所长庄英章先生</as>
CityU	<cityu>选举出现了戏剧性的结果</cityu>

模型	PKU	MSR	AS	CityU
Bakeoff-best	0.950	0.964	-	-
(Zheng, 2013)	0.961	0.974	-	-
(Chen, 2017)	0.943	0.960	-	0.948
(Cai, 2017)	0.958	0.971	0.953	0.956
Bi-LSTM-CRF+多语料	0.960	0.974	0.954	0.962

模型	PKU	MSR	AS	CityU
Bi-LSTM-CRF	0.952	0.973	0.949	0.951
+多语料	0.953	0.973	0.950	0.953
+多语料+标签	0.960	0.974	0.954	0.962

基于异构数据联合训练的中文分词法

Chinese Word Segmentation Based on Joint Training of Heterogeneous Data

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 4

参考文献 16

相关文章 15

Metrics

本文评价

推荐阅读 10

[1]	程晓雅,张雷. 基于深度学习的有遮挡人脸识别方法研究[J]. 电子科技, 2022, 35(1): 35-39.
[2]	赵崇,迟蒙蒙,储聪,张鹏. 导盲犬行走机构运动仿真及其视觉识别算法研究[J]. 电子科技, 2021, 34(9): 66-72.
[3]	马立新,豆晨飞,宋晨灿,杨天笑. 基于特征融合卷积神经网络的绝缘子无损检测[J]. 电子科技, 2021, 34(7): 26-30.
[4]	薛永杰,巨志勇. 基于改进AlexNet的鱼类识别算法[J]. 电子科技, 2021, 34(4): 12-17.
[5]	邵杭,王永雄,秦宇龙. 基于深度学习与显著性的数字图像构图优化[J]. 电子科技, 2021, 34(3): 36-42.
[6]	闫超,孙占全,田恩刚,赵杨洋,范小燕. 基于深度学习的医学图像分割技术研究进展[J]. 电子科技, 2021, 34(2): 7-11.
[7]	程俊华,曾国辉,刘瑾. 基于深度学习的复杂背景图像分类方法研究[J]. 电子科技, 2020, 33(12): 59-66.
[8]	邬倩,吴飞,骆立志. 基于深度学习的人体骨架动作识别[J]. 电子科技, 2020, 33(11): 79-83.
[9]	段雨佳,鞠婷. 基于深度学习的代码审查意见有效性评估[J]. 电子科技, 2020, 33(1): 39-45.
[10]	仝泽友,冯仕民,侯晓晴,丁恩杰. 基于安全帽佩戴检测的矿山人员违规行为研究[J]. 电子科技, 2019, 32(9): 26-31.
[11]	胡少聪. 基于深度学习的人脸识别方法研究[J]. 电子科技, 2019, 32(6): 82-86.
[12]	李荣瑞,施霖,赵薇. 基于卷积神经网络的少数民族头饰识别[J]. 电子科技, 2019, 32(2): 51-55.
[13]	孟昕. 基于深度学习的法律文书识别方法研究[J]. 电子科技, 2019, 32(12): 84-86.
[14]	袁小平,王岗,王晔枫,汪喆远,孙辉. 基于改进卷积神经网络的交通标志识别方法[J]. 电子科技, 2019, 32(11): 28-32.
[15]	史册,裘炅. 基于深度学习机制的火警协议智能分析模型[J]. 电子科技, 2019, 32(1): 80-85.