Filters








563 Hits in 4.0 sec

Shared-Private Bilingual Word Embeddings for Neural Machine Translation

Xuebo Liu, Derek F. Wong, Yang Liu, Lidia S. Chao, Tong Xiao, Jingbo Zhu
2019 Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics  
Word embedding is central to neural machine translation (NMT), which has attracted intensive research interest in recent years.  ...  In this paper, we propose shared-private bilingual word embeddings, which give a closer relationship between the source and target embeddings, and which also reduce the number of model parameters.  ...  Introduction With the introduction of ever more powerful architectures, neural machine translation (NMT) has become the most promising machine translation method (Kalchbrenner and Blunsom, 2013; Sutskever  ... 
doi:10.18653/v1/p19-1352 dblp:conf/acl/LiuWLCXZ19 fatcat:s37v56l3vnefthcfr4ruz3xzpa

Exploiting Japanese-Chinese Cognates with Shared Private Representations for Neural Machine Translation

Zezhong Li, Fuji Ren, Xiao Sun, Degen Huang, Piao Shi
2022 ACM Transactions on Asian and Low-Resource Language Information Processing  
Neural machine translation (NMT) has achieved remarkable progress in the past several years; however, little attention has been paid to MT between Japanese and Chinese who share a large proportion of cognate  ...  with a dynamic gating mechanism, which models the language-specific features on the source side; and (3) an embedding shortcut, which enables the decoder to access the shared private representation with  ...  Second, we learn a shared private embedding with a dynamic gate for the private features on the source side only in each cognate word pair.  ... 
doi:10.1145/3533429 fatcat:y7qfjximirgyppq3uwm54pz6oe

Multi-Source Cross-Lingual Model Transfer: Learning What to Share

Xilun Chen, Ahmed Hassan Awadallah, Hany Hassan, Wei Wang, Claire Cardie
2019 Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics  
Such deep neural models, however, are not applicable to most human languages due to the lack of annotated training data for various NLP tasks.  ...  This enables our model to learn effectively what to share between various languages in the multilingual setup.  ...  BWE baselines rely on Bilingual Word Embeddings (BWEs) and weight sharing for CLTL.  ... 
doi:10.18653/v1/p19-1299 dblp:conf/acl/ChenAHWC19 fatcat:d2l42ilxxzbitjyfl2haontvgq

A Comparative Analysis of Unsupervised Language Adaptation Methods

Gil Rocha, Henrique Lopes Cardoso
2019 Proceedings of the 2nd Workshop on Deep Learning Approaches for Low-Resource NLP (DeepLo 2019)  
In this paper, we explore three recent proposals: Adversarial Training, Sentence Encoder Alignment and Shared-Private Architecture.  ...  Regarding word embeddings, for Chinese we used the pre-trained 50 dimensional Bilingual Word Embeddings (BWE) by Zhou et al. (2016) .  ...  , including fastText embeddings (Joulin et al., 2018; Bojanowski et al., 2017) , Multilingual Unsupervised and Supervised Embeddings (MUSE) , and bilingual word embeddings (BWE) (Zhou et al., 2016) .  ... 
doi:10.18653/v1/d19-6102 dblp:conf/acl-deeplo/RochaC19 fatcat:hr4hxysmqjc2tl5rfy7bgcjebq

Multi-Source Cross-Lingual Model Transfer: Learning What to Share [article]

Xilun Chen, Ahmed Hassan Awadallah, Hany Hassan, Wei Wang, Claire Cardie
2019 arXiv   pre-print
Such deep neural models, however, are not applicable to most human languages due to the lack of annotated training data for various NLP tasks.  ...  This enables our model to learn effectively what to share between various languages in the multilingual setup.  ...  BWE baselines rely on Bilingual Word Embeddings (BWEs) and weight sharing for CLTL.  ... 
arXiv:1810.03552v3 fatcat:mim5ubvnv5ghdkqk3xeyhqfhfm

Focus on the Target's Vocabulary: Masked Label Smoothing for Machine Translation [article]

Liang Chen, Runxin Xu, Baobao Chang
2022 arXiv   pre-print
Label smoothing and vocabulary sharing are two widely used techniques in neural machine translation models.  ...  Our extensive experiments show that MLS consistently yields improvement over original label smoothing on different datasets, including bilingual and multilingual translation from both translation quality  ...  We thank all reviewers for their valuable suggestions for this work.  ... 
arXiv:2203.02889v2 fatcat:w3fkjcshhfcfvc3jhangtzcsa4

Neural Machine Translation for Bilingually Scarce Scenarios: a Deep Multi-Task Learning Approach

Poorya Zaremoodi, Gholamreza Haffari
2018 Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers)  
Neural machine translation requires large amounts of parallel training text to learn a reasonable-quality translation model.  ...  This is particularly inconvenient for language pairs for which enough parallel text is not available.  ...  the members of the JSALT-2017 workshop at CMU, particularly George Foster, Colin Cherry, Patrick Littell, David Mortensen, Graham Neubig, Ji Xin, Daniel Beck, Anna Currey, Vu Hoang, and Gaurav Kumar for  ... 
doi:10.18653/v1/n18-1123 dblp:conf/naacl/ZareMoodiH18 fatcat:qdxygbcanjdrrdl6whwpo3fify

A Multitask-Based Neural Machine Translation Model with Part-of-Speech Tags Integration for Arabic Dialects

Laith H. Baniata, Seyoung Park, Seong-Bae Park
2018 Applied Sciences  
The statistical machine translation for the Arabic language integrates external linguistic resources such as part-of-speech tags.  ...  Machine Translation (NMT) model.  ...  Acknowledgments: This study was supported by the BK21 Plus project (SW Human Resource Development Program for Supporting Smart Life) funded by the Ministry of Education, School of Computer Science and  ... 
doi:10.3390/app8122502 fatcat:pikd3ycupvc23jvhmkphxwbmfq

Neural Machine Translation for Bilingually Scarce Scenarios: A Deep Multi-task Learning Approach [article]

Poorya Zaremoodi, Gholamreza Haffari
2018 arXiv   pre-print
Neural machine translation requires large amounts of parallel training text to learn a reasonable-quality translation model.  ...  This is particularly inconvenient for language pairs for which enough parallel text is not available.  ...  We are very grateful to the workshop members for the insightful discussions and data pre-processing.  ... 
arXiv:1805.04237v1 fatcat:af2ixee74ra4tidacxxj55zonu

Unsupervised Neural Dialect Translation with Commonality and Diversity Modeling [article]

Yu Wan and Baosong Yang and Derek F. Wong and Lidia S. Chao and Haihua Du and Ben C.H. Ao
2019 arXiv   pre-print
Specifically, we leverage pivot-private embedding, layer coordination, as well as parameter sharing to sufficiently model commonality and diversity among source and target, ranging from lexical, through  ...  As a special machine translation task, dialect translation has two main characteristics: 1) lack of parallel training corpus; and 2) possessing similar grammar between two sides of the translation.  ...  Acknowledgements This work was supported in part by the National Natural We thank all the reviewers for their insightful comments.  ... 
arXiv:1912.05134v1 fatcat:e647mz5kqnh2bi7krkcqhu23ae

Unsupervised Neural Dialect Translation with Commonality and Diversity Modeling

Yu Wan, Baosong Yang, Derek F. Wong, Lidia S. Chao, Haihua Du, Ben C.H. Ao
2020 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
Specifically, we leverage pivot-private embedding, layer coordination, as well as parameter sharing to sufficiently model commonality and diversity among source and target, ranging from lexical, through  ...  As a special machine translation task, dialect translation has two main characteristics: 1) lack of parallel training corpus; and 2) possessing similar grammar between two sides of the translation.  ...  Acknowledgements This work was supported in part by the National Natural We thank all the reviewers for their insightful comments.  ... 
doi:10.1609/aaai.v34i05.6448 fatcat:sjiseetnirbhtk3owx5yokpztm

English–Welsh Cross-Lingual Embeddings

Luis Espinosa-Anke, Geraint Palmer, Padraig Corcoran, Maxim Filimonov, Irena Spasić, Dawn Knight
2021 Applied Sciences  
Cross-lingual embeddings are vector space representations where word translations tend to be co-located.  ...  We used a bilingual dictionary to frame the problem of learning bilingual mappings as a supervised machine learning task, where a word vector space is first learned independently on a monolingual corpus  ...  Acknowledgments: The research on which this article is based was funded by the Welsh Government as part of the "Learning English-Welsh bilingual embeddings and applications in text categorisation" project  ... 
doi:10.3390/app11146541 fatcat:aht2dnh6xje4ldtw5zhdimmkea

Cross-lingual Emotion Intensity Prediction [article]

Irean Navas Alejo, Toni Badia, Jeremy Barnes
2020 arXiv   pre-print
We compare six cross-lingual approaches, e.g., machine translation and cross-lingual embeddings, which have varying requirements for parallel data – from millions of parallel sentences to completely unsupervised  ...  Consequently, we explore cross-lingual transfer approaches for fine-grained emotion detection in Spanish and Catalan tweets.  ...  ) and then use VecMap (Artetxe et al., 2017) to learn an orthogonal projection of the word embeddings to a joint shared embedding space using a small bilingual lexicon 4 as supervision (5749 and 5310  ... 
arXiv:2004.04103v2 fatcat:gcxekez4d5aodkaaie3uxdyqce

PROMT Systems for WMT 2018 Shared Translation Task

Alexander Molchanov
2018 Proceedings of the Third Conference on Machine Translation: Shared Task Papers  
This paper describes the PROMT submissions for the WMT 2018 Shared News Translation Task. This year we participated only in the English-Russian language pair.  ...  We also submitted pure rule-based translation (RBMT) for contrast. We show competitive results with both primary submissions which significantly outperform the RBMT baseline.  ...  Introduction This paper provides an overview of the PROMT submissions for the WMT 2018 Shared News Translation Task. This year we participate with neural MT systems for the first time.  ... 
doi:10.18653/v1/w18-6420 dblp:conf/wmt/Molchanov18 fatcat:jl2lzamtxnaj5drhn54bumih7y

XNLI: Evaluating Cross-lingual Sentence Representations

Alexis Conneau, Ruty Rinott, Guillaume Lample, Adina Williams, Samuel Bowman, Holger Schwenk, Veselin Stoyanov
2018 Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing  
In addition, we provide several baselines for multilingual sentence understanding, including two based on machine translation systems, and two that use parallel data to train aligned multilingual bag-of-words  ...  We find that XNLI represents a practical and challenging evaluation suite, and that directly translating the test data yields the best performance among available baselines.  ...  Cross-lingual embeddings also provide an efficient mechanism to bootstrap neural machine translation (NMT) systems for low-resource language pairs, which is critical in the case of unsupervised machine  ... 
doi:10.18653/v1/d18-1269 dblp:conf/emnlp/ConneauRLWBSS18 fatcat:hnfqpqitfjduxgs3cbgvosuey4
« Previous Showing results 1 — 15 out of 563 results