[1]武子玉,侯宏旭*,白天罡,等.基于CSGAN的多模型融合蒙汉神经机器翻译研究[J].江西师范大学学报(自然科学版),2020,(02):153-159.[doi:10.16357/j.cnki.issn1000-5862.2020.02.07]
　WU Ziyu,HOU Hongxu*,BAI Tiangang,et al.The Research on Multi-Model Fusion for Mongolian-Chinese Neural Machine Translation Based on CSGAN[J].Journal of Jiangxi Normal University:Natural Science Edition,2020,(02):153-159.[doi:10.16357/j.cnki.issn1000-5862.2020.02.07]
点击复制

基于CSGAN的多模型融合蒙汉神经机器翻译研究()
分享到：

《江西师范大学学报》（自然科学版）[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2020年02期
页码:
153-159
栏目:
出版日期:
2020-04-10

文章信息/Info

Title:
The Research on Multi-Model Fusion for Mongolian-Chinese Neural Machine Translation Based on CSGAN
文章编号:
1000-5862(2020)02-0153-07
作者:
武子玉1; 侯宏旭2*; 白天罡2; 吉亚图2; 乌尼尔2; 郭紫月2; 王雪姣2; 孙 硕2
1.内蒙古大学计算机学院,内蒙古 呼和浩特 010021; 2.内蒙古自治区蒙古文信息处理技术重点实验室,内蒙古 呼和浩特 010021
Author(s):
WU Ziyu1; HOU Hongxu2*; BAI Tiangang2; JI Yatu2; WU Nier2; GUO Ziyue2; WANG Xuejiao2; SUN Shuo2
1.College of Computer Science,Inner Mongolia University, Hohhot Inner Mongolia 010021,China; 2.Inner Mongolia A.R. Key Laboratory of Mongolian Information Processing Technology,Inner Mongolia University,Hohhot Inner Mongolia 010021,China
关键词:
蒙汉机器翻译;  数据稀疏;  系统融合;  命名实体
Keywords:
Mongolian-Chinese machine translation;  data sparsity;  fusion system;  named entity
分类号:
TP 391.1
DOI:
10.16357/j.cnki.issn1000-5862.2020.02.07
文献标志码:
A
摘要:
由于低资源语料稀少而导致的语义捕获不充分现象已成为影响机器翻译质量的主要因素.为此,该文在预处理的基础上利用CNN和门控机制来改进Transformer模型,通过对抗训练的方式来引导模型参数的优化,同时通过加入命名实体识别来提高模型对实体的翻译性能.此外,通过多模型融合的方式将来自多个机器翻译的输出经过改进、重组、合并转变为一个单一的改进的翻译结果.通过3组对比实验表明,该方法优于基准方法.
Abstract:
The phenomenon of insufficient semantic capture due to the scarcity of low-resource corpus has become a major factor affecting the quality of machine translation.Therefore,based on the pretreatment,the paper improves the transformer model by using CNN and gating mechanism and guides the optimization of model parameters using confrontation training.At the same time,named entity recognition is added to improve the translation performance of model entities.Reorganize and merge the output from multiple machine translations into a single improved translation result through multi-model fusion.The Mongolian-Chinese translation experiments show that the proposed method is superior to the benchmark method.

参考文献/References:

[1] Sennrich R,Haddow B,Birch A.Neural machine translation of rare words with subword units[EB/OL].[2019-03-16].https://arxiv.org/abs/1508.07909.
[2] 清格尔泰.蒙古语语法[M].呼和浩特:内蒙古人民出版社,1992.
[3] Le H Q,Nguyen T M,Vu S T,et al.D3NER:Biomedical named entity recognition using CRF-biLSTM improved with fine-tuned embeddings of various linguistic information[J].Bioinformatics,2018,34(20):32-36.
[4] Gehring J,Auli M,Grangier D,et al.Convolutional sequence to sequence learning[EB/OL].[2019-03-19].https://arxiv.org/abs/1705.03122.
[5] Gehring J,Auli M,Grangier D,et al.A convolutional encoder model for neural machine translation[EB/OL].[2019-03-19].https://arxiv.org/abs/1611.02344.
[6] Vaswani A,Shazeer N,Parmar N,et al.Attention is all you need[EB/OL].[2019-03-13].https://arxiv.org/abs/1706.03762.
[7] He Kaiming,Zhang Xiangyu,Ren Shaoqing,et al.Deep residual learning for image recognition[EB/OL].[2019-04-13].https://arxiv.org/abs/1512.03385.
[8] Ba J L,Kiros J R,Hinton G E.Layer normalization[EB/OL].[2019-04-18].arXiv preprint arXiv:1607.06450,2016.
[9] Glorot X,Bordes A,Bengio Y.Deep sparse rectifier neural networks[EB/OL].[2019-04-22].https://www.researchgate.net/publication/215616967_Deep_Sparse_Rectifier_Neural_Networks.
[10] Dauphin Y N,Fan A,Auli M,et al.Language modeling with gated linear units[EB/OL].[2019-04-25].https://arxiv.org/abs/1612.08083.
[11] He Kaiming,Zhang Xiangyu,Ren Shaoqing,et al.Deep residual learning for image recognition[EB/OL].[2019-04-29].https://arxiv.org/abs/1512.03385.
[12] Mirza M,Osindero S.Conditional generative adversarial nets[EB/OL].[2019-03-29].https://arxiv.org/abs/1411.1784.
[13] Bangalore S,Bordel G,Riccardi G.Computing consensus translation from multiple machine translation systems[EB/OL].[2019-03-12].http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.69.4528.
[14] Heafield K,Lavie A.Combining machine translation output with open source:the carnegie mellon multi-engine machine translation scheme[J].Prague Bulletin of Mathematical Linguistics,2010,93(1):27-36.
[15] Rosti A V I,Matsoukas S,Schwartz R M.Improved word-level system combination for machine translation[C].Proceedings of the 45th Annual Meeting of the Association for Computational Linguistics,Prague:DBLP,2007:312-319.
[16] Banerjee S,Lavie A.METEOR:an automatic metric for MT evaluation with improved correlation with human judgments[EB/OL].[2019-03-02].
[17] Hieu Hoang,Alexandra Birch,Chris Callison-burch,et al.Moses:open source toolkit for statistical machine translation[EB/OL].[2019-05-02].https://dl.acm.org/citation.cfm?doid=1557769.1557821.
[18] Snover M,Dorr B,Schwartz R,et al.A study of translation edit rate with targeted human annotation[EB/OL].[2019-05-05].https://www.researchgate.net/publication/228668187_A_study_of_translation_edit_rate_with_targeted_human_annotation.

备注/Memo

备注/Memo:
收稿日期:2019-09-02
基金项目:内蒙古自然科学基金(2018MS06005)和内蒙古自治区科技成果转化“蒙古文机器翻译与辅助翻译云平台建设与推广”(2019CG028)资助项目.
通信作者:侯宏旭(1972-),男,山东济南人,教授,博士,主要从事自然语言处理、信息检索研究.E-mail:cshhx@imu.edu.cn
更新日期/Last Update: 2020-04-10