Filters








1,357 Hits in 4.9 sec

Multi-Task Learning for Improved Discriminative Training in SMT

Patrick Simianer, Stefan Riezler
2013 Conference on Machine Translation  
Multi-task learning has been shown to be effective in various applications, including discriminative SMT.  ...  We find that both versions of multi-task learning improve equally well over independent and pooled baselines, and gain nearly 2 BLEU points over standard MERT tuning.  ...  Acknowledgments The research presented in this paper was supported in part by DFG grant "Cross-language Learningto-Rank for Patent Retrieval".  ... 
dblp:conf/wmt/SimianerR13 fatcat:wijpgkklirem5dmelf2sagwv2y

Multi-Task Minimum Error Rate Training for SMT

Patrick Simianer, Katharina Wäschle, Stefan Riezler
2011 Prague Bulletin of Mathematical Linguistics  
Multi-Task Minimum Error Rate Training for SMT We present experiments on multi-task learning for discriminative training in statistical machine translation (SMT), extending standard minimum-error-rate  ...  training (MERT) by techniques that take advantage of the similarity of related tasks.  ...  Our goal is to investigate how state-of-the-art multi-task learning techniques for linear classifiers can be applied to standard discriminative training for SMT.  ... 
doi:10.2478/v10108-011-0015-0 fatcat:eepkepl6lnhkld2bizwytesrn4

Joint Feature Selection in Distributed Stochastic Learning for Large-Scale Discriminative Training in SMT

Patrick Simianer, Stefan Riezler, Chris Dyer
2012 Annual Meeting of the Association for Computational Linguistics  
With a few exceptions, discriminative training in statistical machine translation (SMT) has been content with tuning weights for large feature sets on small development data.  ...  We present experiments on learning on 1.5 million training sentences, and show significant improvements over tuning discriminative models on small development sets.  ...  Induction of Multi-Nonterminal Grammars for SMT" from Google, Inc.  ... 
dblp:conf/acl/SimianerRD12 fatcat:fcubecyv4bhy7op2pdnli55iym

Multi-Domain Adaptation for SMT Using Multi-Task Learning

Lei Cui, Xilun Chen, Dongdong Zhang, Shujie Liu, Mu Li, Ming Zhou
2013 Conference on Empirical Methods in Natural Language Processing  
In this paper, we propose a novel multi-domain adaptation approach for SMT using Multi-Task Learning (MTL), with in-domain models tailored for each specific domain and a general-domain model shared by  ...  Domain adaptation for SMT usually adapts models to an individual specific domain.  ...  Acknowledgments We are especially grateful to Nan Yang, Yajuan Duan, Hong Sun and Danran Chen for the helpful discussions. We also thank the anonymous reviewers for their insightful comments.  ... 
dblp:conf/emnlp/CuiCZLLZ13 fatcat:lxyq4vgdj5h3jc4tg5xvtofn7m

A Survey of Domain Adaptation for Machine Translation

Chenhui Chu, Rui Wang
2020 Journal of Information Processing  
Neural machine translation (NMT) is a deep learning based approach for machine translation, which outperforms traditional statistical machine translation (SMT) and yields the state-of-the-art translation  ...  Because of the current dominance of NMT in MT research, we give a brief review of domain adaptation for SMT, but put most of our effort into the survey of domain adaptation for NMT.  ...  Acknowledgments This work was supported by Grant-in-Aid for Young Scientists #19K20343, JSPS and Microsoft Research Asia Collaborative Research Grant. We are very indebted to Dr.  ... 
doi:10.2197/ipsjjip.28.413 fatcat:eeboqsm6rfbu7hd6dr4q6u4o3e

Improving Statistical Machine Translation Using Word Sense Disambiguation

Marine Carpuat, Dekai Wu
2007 Conference on Empirical Methods in Natural Language Processing  
In this paper, we address this problem by investigating a new strategy for integrating WSD into an SMT system, that performs fully phrasal multi-word disambiguation.  ...  We show for the first time that incorporating the predictions of a word sense disambiguation system within a typical phrase-based statistical machine translation (SMT) model consistently improves translation  ...  the WSD system for multi-word phrase-based SMT.  ... 
dblp:conf/emnlp/CarpuatW07 fatcat:jy7lyu6cprevjdspjexangw7g4

The HDU Discriminative SMT System for Constrained Data PatentMT at NTCIR10

Patrick Simianer, Gesa Stupperich, Laura Jehl, Katharina Wäschle, Artem Sokolov, Stefan Riezler
2013 NTCIR Conference on Evaluation of Information Access Technologies  
Our goal is to address the twofold nature of patents by exploiting the repetitive nature of patents through feature sharing in a multi-task learning setup (used in the Japaneseto-English translation subtask  ...  The core system used in both subtasks is a combination of hierarchical phrase-based translation and discriminative training using either large feature sets and 1/ 2 regularization (for Japanese-to-English  ...  Acknowledgments The work presented in this paper was supported in part by DFG grant "Cross-language Learning-to-Rank for Patent Retrieval".  ... 
dblp:conf/ntcir/SimianerSJWSR13 fatcat:kw35kai7fzfjbkalrjql2gy3r4

Bagging and Boosting statistical machine translation systems

Tong Xiao, Jingbo Zhu, Tongran Liu
2013 Artificial Intelligence  
The situation is even more severe when the learning algorithm is prone to different local optimal and/or no sufficient training data is provided.  ...  ., different features weights in the log-linear model) can be learned, and the "best" system is in general selected according to some criteria.  ...  [31] recast re-ranking as a multi-task learning problem and investigated methods to ease discriminative training when a large number of features are involved.  ... 
doi:10.1016/j.artint.2012.11.005 fatcat:vy3igxvjz5cc5dqqg6vdihcd3u

A Survey of Domain Adaptation for Neural Machine Translation [article]

Chenhui Chu, Rui Wang
2018 arXiv   pre-print
Neural machine translation (NMT) is a deep learning based approach for machine translation, which yields the state-of-the-art translation performance in scenarios where large-scale parallel corpora are  ...  In this paper, we give a comprehensive survey of the state-of-the-art domain adaptation techniques for NMT.  ...  Acknowledgement This work was supported by Grant-in-Aid for Research Activity Start-up #17H06822, JSPS. We are very appreciated to Dr.  ... 
arXiv:1806.00258v1 fatcat:jherla25kbalhe3b3k2gzo6qba

Learning to Discriminate Noises for Incorporating External Information in Neural Machine Translation [article]

Zaixiang Zheng, Shujian Huang, Zewei Sun, Rongxiang Weng, Xin-Yu Dai, Jiajun Chen
2018 arXiv   pre-print
Experimental results in various real-world scenarios, language pairs, and neural architectures indicate that discriminating noises contributes to significant improvements in translation quality by being  ...  Accordingly, we propose a general framework that learns to jointly discriminate both the global and local noises, so that the external information could be better leveraged.  ...  Note that in real-world settings, we use the same trained models from our main experiments without task-specific tuning for the different given datasets.  ... 
arXiv:1810.10317v3 fatcat:2rj3gigjrvbwpft7tfcl2og5ny

Bagging-based System Combination for Domain Adaption

Linfeng Song, Haitao Mi, Yajuan Lü, Qun Liu
2011 Machine Translation Summit  
Domain adaptation plays an important role in multi-domain SMT.  ...  Conventional approaches usually resort to statistical classifiers, but they require annotated monolingual data in different domains, which may not be available in some cases.  ...  We are grateful to the anonymous reviewers for their valuable comments.  ... 
dblp:conf/mtsummit/SongML011 fatcat:ajolhoazdveltj5o6uw3skcaxq

Structural and Topical Dimensions in Multi-Task Patent Translation

Katharina Wäschle, Stefan Riezler
2012 Conference of the European Chapter of the Association for Computational Linguistics  
We study multitask learning techniques that exploit commonalities between tasks by mixtures of translation models or by multi-task metaparameter tuning.  ...  We find small but significant gains over task-specific training by techniques that model commonalities through shared parameters.  ...  Acknowledgments This work was supported in part by DFG grant "Cross-language Learning-to-Rank for Patent Retrieval".  ... 
dblp:conf/eacl/WaeschleR12 fatcat:7wfugeq5m5fm3n474bs5vkix3u

Discriminative Phrase-Based Models for Arabic Machine Translation

Cristina España-Bonet, Jesús Giménez, Lluís Màrquez
2009 ACM Transactions on Asian Language Information Processing  
These classifiers are integrated into the translation system so that the global task gets benefits from the discriminative learning.  ...  Although using discriminative learning methods for SMT can be useful for any language pair, those source languages with especially ambiguous semantics where words tend to have a larger number of lexical  ...  FULL TRANSLATION TASK In the following, we investigate whether the improvement obtained for the local task of phrase selection has a positive repercussion on the global translation task.  ... 
doi:10.1145/1644879.1644882 fatcat:azgqvwm4sze6lhzide4k3lmfiy

Burst2Vec: An Adversarial Multi-Task Approach for Predicting Emotion, Age, and Origin from Vocal Bursts [article]

Atijit Anuchitanukul, Lucia Specia
2022 arXiv   pre-print
We present Burst2Vec, our multi-task learning approach to predict emotion, age, and origin (i.e., native country/language) from vocal bursts.  ...  Our models achieve a relative 30 % performance gain over baselines using pre-extracted features and score the highest amongst all participants in the ICML ExVo 2022 Multi-Task Challenge.  ...  for multi-task learning.  ... 
arXiv:2206.12469v1 fatcat:5pawnwquarbqbdf56knfkhc5um

Discriminative adaptation of continuous space translation models

Quoc-Khanh Do, Alexandre Allauzen, François Yvon
2014 International Workshop on Spoken Language Translation  
In our experiments, the baseline out-of-domain SMT system is initially trained for the WMT News translation task, and the CSTM is to be adapted to the lecture translation task as defined by IWSLT evaluation  ...  We consider the following practical situation: given a large scale, stateof-the-art SMT system containing a CSTM, the task is to adapt the CSTM to a new domain using a (relatively) small in-domain parallel  ...  Our work owes much to recent contributions in discriminative training and tuning of SMT systems.  ... 
dblp:conf/iwslt/DoAY14 fatcat:otq632i4vjf7jfkiddegubdgpm
« Previous Showing results 1 — 15 out of 1,357 results