|本期目录/Table of Contents|

[1]谭 新,邝少辉,张龙印,等.融入汉字笔画序列的神经机器翻译[J].厦门大学学报(自然科学版),2019,58(02):164-169.[doi:10.6043/j.issn.0438-0479.201811023]
 TAN Xin,KUANG Shaohui,ZHANG Longyin,et al.Integration of Chinese character stroke sequence into neural machine translation[J].Journal of Xiamen University(Natural Science),2019,58(02):164-169.[doi:10.6043/j.issn.0438-0479.201811023]
点击复制

融入汉字笔画序列的神经机器翻译(PDF/HTML)
分享到:

《厦门大学学报(自然科学版)》[ISSN:0438-0479/CN:35-1070/N]

卷:
58卷
期数:
2019年02期
页码:
164-169
栏目:
机器翻译模型
出版日期:
2019-03-27

文章信息/Info

Title:
Integration of Chinese character stroke sequence into neural machine translation
文章编号:
0438-0479(2019)02-0164-06
作者:
谭 新邝少辉张龙印熊德意*
苏州大学计算机科学与技术学院,江苏 苏州 215006
Author(s):
TAN XinKUANG ShaohuiZHANG LongyinXIONG Deyi*
School of Computer Science and Technology,Soochow University,Suzhou 215006,China
关键词:
神经机器翻译 汉字笔画序列 注意力机制
Keywords:
neural machine translation Chinese character stroke sequence attention mechanism
分类号:
TP 391
DOI:
10.6043/j.issn.0438-0479.201811023
文献标志码:
A
摘要:
神经机器翻译(NMT)因其在多个语言对上的翻译效果都远超传统的统计机器翻译(SMT)而逐渐成为机器翻译方向的主流.然而,这种NMT系统在将向量化的词语作为输入时只考虑了词语整体的语义信息,忽略了构成词语的汉字本身所包含的信息.为此,针对汉字给出了一种融入汉字笔画序列的NMT系统.该系统在将词语的词向量作为输入的同时又将向量化的汉字笔画序列作为额外输入,既考虑了中文词语整体的语义信息,又考虑了构成词语的汉字本身的内部语义信息和外部形态信息.实验结果表明,提出的融入了汉字笔画序列的NMT系统更加有效,其翻译结果更加准确流畅,与传统的NMT系统相比机器双语互译评估(BLEU)值能够提高1.21个百分点.
Abstract:
Neural machine translation(NMT)has gradually become the mainstream of machine translation because it surpasses statistical machine translation(SMT)in a variety of language pairs.However,conventional NMT systems merely take word embeddings into account,ignoring the information contained in Chinese characters that constitute words.Therefore,this paper proposes a new approach that integrates Chinese character stroke sequence information into NMT system.Our proposed method considers both word embeddings for Chinese words and Chinese character stroke sequence information into NMT system.We consider not only the semantic information of each Chinese word,but also the internal semantic information and external form information of Chinese characters that form these words.Experimental results show that our proposed method,which integrates Chinese character stroke sequence information into NMT system,is more effective and makes translation results more accurate.Compared with traditional NMT system,our proposed method can achieve an improvement of BLEU value by 1.21 percentage points.

参考文献/References:

[1] SUTSKEVER I,VINYALS O,LE Q V.Sequence to sequence learning with neural networks[C]∥Advances in Neural Information Processing Systems.Montreal:NIPS,2014:3104-3112.
[2] CHO K,VAN MERRI?NBOER B,GULCEHRE C,et al.Learning phrase representations using RNN encoder-decoder for statistical machine translation[EB/OL].[2018-11-11].https:∥arxiv.org/pdf/1406.1078.
[3] CHO K,VAN MERRI?NBOER B,BAHDANAU D,et al.On the properties of neural machine translation:encoder-decoder approaches[EB/OL].[2018-11-11].https:∥arxiv.org/pdf/1409.1259.
[4] CHIANG D.A hierarchical phrase-based model for statistical machine translation[C]∥Proceedings of the 43rd Annual Meeting on Association for Computational Linguistics.Arbor:ACL,2005:263-270.
[5] COTTERELL R,SCHüTZE H.Morphological word-embeddings[C]∥Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies.Colorado:NAACL,2015:1287-1292.
[6] BOJANOWSKI P,GRAVE E,JOULIN A,et al.Enriching word vectors with subword information[EB/OL].[2018-11-11].https:∥arxiv.org/pdf/1607.04606.
[7] KUANG S,HAN L.Apply Chinese radicals into neural machine translation:deeper than character level[EB/OL].[2018-11-11].https:∥arxiv.org/pdf/1805.01565.
[8] BAHDANAU D,CHO K,BENGIO Y.Neural machine translation by jointly learning to align and translate[EB/OL].[2018-11-11].https:∥arxiv.org/pdf/1409.0473.
[9] CAO S,LU W,ZHOU J,et al.Cw2vec:learning Chinese word embedding with stroke n-gram information[C]∥The 32nd AAAI Conference on Artificial Intelligence.New Orleans:AAAI,2018:5053-5061.
[10] SRIVASTAVA N,HINTON G,KRIZHEVSKY A,et al.Dropout:a simple way to prevent neural networks from overfitting[J].The Journal of Machine Learning Research,2014,15(1):1929-1958.
[11] ZEILER M D.ADADELTA:an adaptive learning rate method[EB/OL].[2018-11-11].https:∥arxiv.org/pdf/1212.5701.
[12] PAPINENI K,ROUKOS S,WARD T,et al.BLEU:a method for automatic evaluation of machine translation[C]∥Proceedings of the 40th Annual Meeting on Association for Computational Linguistics.Pennsylvania:ACL,2002:311-318.[1] SUTSKEVER I,VINYALS O,LE Q V.Sequence to sequence learning with neural networks[C]∥Advances in Neural Information Processing Systems.Montreal:NIPS,2014:3104-3112.
[2] CHO K,VAN MERRI?NBOER B,GULCEHRE C,et al.Learning phrase representations using RNN encoder-decoder for statistical machine translation[EB/OL].[2018-11-11].https:∥arxiv.org/pdf/1406.1078.
[3] CHO K,VAN MERRI?NBOER B,BAHDANAU D,et al.On the properties of neural machine translation:encoder-decoder approaches[EB/OL].[2018-11-11].https:∥arxiv.org/pdf/1409.1259.
[4] CHIANG D.A hierarchical phrase-based model for statistical machine translation[C]∥Proceedings of the 43rd Annual Meeting on Association for Computational Linguistics.Arbor:ACL,2005:263-270.
[5] COTTERELL R,SCHüTZE H.Morphological word-embeddings[C]∥Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies.Colorado:NAACL,2015:1287-1292.
[6] BOJANOWSKI P,GRAVE E,JOULIN A,et al.Enriching word vectors with subword information[EB/OL].[2018-11-11].https:∥arxiv.org/pdf/1607.04606.
[7] KUANG S,HAN L.Apply Chinese radicals into neural machine translation:deeper than character level[EB/OL].[2018-11-11].https:∥arxiv.org/pdf/1805.01565.
[8] BAHDANAU D,CHO K,BENGIO Y.Neural machine translation by jointly learning to align and translate[EB/OL].[2018-11-11].https:∥arxiv.org/pdf/1409.0473.
[9] CAO S,LU W,ZHOU J,et al.Cw2vec:learning Chinese word embedding with stroke n-gram information[C]∥The 32nd AAAI Conference on Artificial Intelligence.New Orleans:AAAI,2018:5053-5061.
[10] SRIVASTAVA N,HINTON G,KRIZHEVSKY A,et al.Dropout:a simple way to prevent neural networks from overfitting[J].The Journal of Machine Learning Research,2014,15(1):1929-1958.
[11] ZEILER M D.ADADELTA:an adaptive learning rate method[EB/OL].[2018-11-11].https:∥arxiv.org/pdf/1212.5701.
[12] PAPINENI K,ROUKOS S,WARD T,et al.BLEU:a method for automatic evaluation of machine translation[C]∥Proceedings of the 40th Annual Meeting on Association for Computational Linguistics.Pennsylvania:ACL,2002:311-318.

备注/Memo

备注/Memo:
收稿日期:2018-11-12 录用日期:2018-12-05
基金项目:国家自然科学基金(61622209,61861130364)
*通信作者:dyxiong@suda.edu.cn
更新日期/Last Update: 1900-01-01