[1]施寒瑜,曲维光,魏庭新,等.基于组合深度模型的现代汉语数量名短语识别[J].南京师大学报(自然科学版),2022,45(01):127-135.[doi:10.3969/j.issn.1001-4616.2022.01.018]
　Shi Hanyu,Qu Weiguang,Wei Tingxin,et al.Quantity Noun Phrase Structure Recognition Based on Combined Deep Learning Model[J].Journal of Nanjing Normal University(Natural Science Edition),2022,45(01):127-135.[doi:10.3969/j.issn.1001-4616.2022.01.018]
点击复制

基于组合深度模型的现代汉语数量名短语识别()
分享到：

《南京师大学报（自然科学版）》[ISSN:1001-4616/CN:32-1239/N]

卷:
第45卷
期数:
2022年01期
页码:
127-135
栏目:
·计算机科学与技术·
出版日期:
2022-03-15

文章信息/Info

Title:
Quantity Noun Phrase Structure Recognition Based on Combined Deep Learning Model
文章编号:
1001-4616(2022)01-0127-09
作者:
施寒瑜1; 曲维光1; 2; 魏庭新2; 3; 周俊生1; 顾彦慧1
(1.南京师范大学计算机与电子信息学院/人工智能学院,江苏 南京 210023)(2.南京师范大学文学院,江苏 南京 210097)(3.南京师范大学国际文化教育学院,江苏 南京 210097)
Author(s):
Shi Hanyu1; Qu Weiguang1; 2; Wei Tingxin2; 3; Zhou Junsheng1; Gu Yanhui1
(1.School of Computer and Electronic Information/School of Artificial Intelligence,Nanjing Normal University,Nanjing 210023,China)(2.School of Chinese Language and Literature,Nanjing Normal University,Nanjing 210097,China)(3.International College for Chinese Studies,Nanjing Normal University,Nanjing 210097,China)
关键词:
数量名短语识别; BERT; Lattice LSTM;  CRF
Keywords:
the recognition of quantity noun phrases; BERT; lattice LSTM; CRF
分类号:
TP391
DOI:
10.3969/j.issn.1001-4616.2022.01.018
文献标志码:
A
摘要:
数量名短语的识别是识别由数量短语修饰的名词短语左右边界的研究. 以往研究中,基于统计学习模型的数量短语识别方法依赖人工特征,需要通过专家知识构建知识库来实现对“数词+量词”短语的识别. 本文在以往研究基础上纳入“名词”形成“数词+量词+名词”等八类数量名短语,并采用深度学习方法解决这一边界识别任务. 通过BERT模型对原始文本进行上下文特征表示,利用Lattice LSTM模型字词结合的思想将标准分词作为软特征融入文本字符级的特征表示中,最后通过CRF全局约束识别数量名短语边界. 实验结果表明,本文方法在AMR语料上达到较优结果,精确率、召回率、F1值分别为80.83%,89.78%,85.07%.
Abstract:
The research on recognition of quantity noun phrases is the identity of the left and right boundaries of quantity noun phrases. In previous studies,this task focuses on the recognition of quantity phrase and relies on artifical features which are constructed by experts based on statistical learning models. In this paper,we aim at the recognition of quantity noun phrases which have 8 subtypes and propose a neural network model to address the issue. Firstly,BERT is used to represent the contextual features of the original text. Then,the standard word segmentation is incorporated into the feature representation of the text character level as a soft feature by using the idea of Lattice LSTM model. Finally,the left and right boundaries of the“quantity noun phrase”are identified by the CRF global constraint. The experimental results show that this method achieves the better results and the precision,recall and F1 value reaches 80.83%,89.78%,85.07% respectively in the corpus of CAMR.

参考文献/References:

[1] 黎锦熙. 论现代汉语中的量词[M]. 北京:商务印书馆,1978.
[2]白晓革,李义杰. 数量短语的构成模式及其识别[C]//第三届HNC与语言学研究学术研讨会论文集,北京,2005:171-178.
[3]张玲,熊文,李义杰,等. 基于知识库的现代汉语数量短语的识别[C]//第七届中文信息处理国际会议论文集,武汉,2007:295-299.
[4]熊文,张玲. 一种基于规则不依赖于分词的中文数量短语的识别[C]//第七届中文信息处理国际会议论文集,武汉,2007:36-40.
[5]方芳,李斌. 基于语料库的数量名短语识别[C]//第三届学生计算语言学研讨会论文集,沈阳,2006:331-337.
[6]PINHEIRO P H O,COLLOBERT R. Recurrent convolutional neural networks for scene parsing[EB/OL].(2013-06-12)
[2019-11-4]. //https://arxiv.org/abs/1306.2795.
[7]HUANG Z H,XU W,YU K. Bidirectional LSTM-CRF models for sequence tagging[EB/OL].(2015-08-09)
[2019-11-4]. https://arxiv.org/abs/1508.01991.
[8]CHIU J P C,NICHOLS E. Named entity recognition with bidirectional LSTM-CNNs[J]. Transaction of the association of computational linguistics,2016(4):357-370.
[9]曲维光,周俊生,吴晓东,等. 自然语言句子抽象语义表示AMR研究综述[J]. 数据采集与处理,2017,32(1):26-36.
[10]李斌,闻媛,宋丽,等. 融合概念对齐信息的中文AMR语料库的构建[J]. 中文信息学报,2017,31(6):93-102.
[11]PETERS M E,NEUMANN M,LYYER M,et al. Deep contextualized word representations[C]//Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies,New Orleans,Louisiana,United States of America. 2018:2227-2237.
[12]DEVLIN J,CHANG M W,LEE K,et al. BERT:Pre-training of deep bidirectional transformers for language understanding[C]//Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies,2019(1):4171-4186.
[13]VASWANI A,SHAZEER N,PARMAR N,et al. Attention is all you need[C]//Advances in Neural Information Processing Systems 30:Annual Conference on Neural Information Processing Systems,2017:5998-6008.
[14]HOCHREITER S,SCHMIDHUBER J. Long short-term memory[J]. Neural computation,1997,9(8):1735-1780.
[15]ZHANG Y,YANG J. Chinese NER using lattice LSTM[C]//Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics,2018:1554-1564.
[16]LAFFERTY J D,MCCALLUM A,PEREIRA F C N. Conditional random fields:probabilistic models for segmenting and labeling sequence data[C]//Proceedings of the Eighteenth International Conference on Machine Learning,2001:282-289.
[17]RATINOV L,ROTH D. Design challenges and misconceptions in named entity recognition[C]//Proceedings of the Thirteenth Conference on Computational Natural Language Learning,2009:147-155.
[18]SRIVASTAVA N,HINTON G E,KRIZHEVSKY A,et al. Dropout:a simple way to prevent neural networks from overfitting[J]. Journay machine learning research,2014,15(1):1929-1958.
[19]ZEILER M D. ADADELTA:an adaptive learning rate method[EB/OL].(2012-12-22)
[2019-11-4]. //https://arxiv.org/abs/1212.5701.
[20]KINGMA D P,BA J. Adam:a method for stochastic optimization[C]//3rd International Conference on Learning Representations,2015:arXiv:1412.6980.
[21]DAUPHIN Y N,VRIES H D,BENGIO Y. Equilibrated adaptive learning rates for non-convex optimization[C]//Advances in Neural Information Processing Systems 28:Annual Conference on Neural Information Processing Systems,2015:1504-1512.

备注/Memo

备注/Memo:
收稿日期:2020-12-26.
基金项目:国家自然科学基金项目(61772278、61472191)、国家社科基金项目(21&ZD288、18BYY127).
通讯作者:曲维光,博士,教授,研究方向:自然语言处理. E-mail:wgqu_nj@163.com
更新日期/Last Update: 1900-01-01