Edinburgh s Neural Machine Translation Systems
|
|
|
- Edward Page
- 5 years ago
- Views:
Transcription
1 Edinburgh s Neural Machine Translation Systems Barry Haddow University of Edinburgh October 27, 2016 Barry Haddow Edinburgh s NMT Systems 1 / 20
2 Collaborators Rico Sennrich Alexandra Birch Barry Haddow Edinburgh s NMT Systems 1 / 20
3 Edinburgh s WMT Results Over the Years BLEU on newstest2013 (EN DE) phrase-based SMT syntax-based SMT neural MT Barry Haddow Edinburgh s NMT Systems 2 / 20
4 Edinburgh s WMT Results Over the Years BLEU on newstest2013 (EN DE) phrase-based SMT syntax-based SMT neural MT Barry Haddow Edinburgh s NMT Systems 2 / 20
5 Edinburgh s WMT Results Over the Years BLEU on newstest2013 (EN DE) phrase-based SMT syntax-based SMT neural MT Barry Haddow Edinburgh s NMT Systems 2 / 20
6 Neural Machine Translation: Encoder-Decoder-with-Attention [Image: Philipp Koehn] Barry Haddow Edinburgh s NMT Systems 3 / 20
7 Neural versus Phrase-based MT phrase-based SMT Learn segment-segment correspondences from bitext training is multistage pipeline of heuristics strong independence assumptions fixed trade-off between features Barry Haddow Edinburgh s NMT Systems 4 / 20
8 Neural versus Phrase-based MT phrase-based SMT Learn segment-segment correspondences from bitext training is multistage pipeline of heuristics strong independence assumptions fixed trade-off between features neural MT Learn mathematical function on vectors from bitext end-to-end trained model output conditioned on full source text and target history non-linear dependence on information sources Barry Haddow Edinburgh s NMT Systems 4 / 20
9 Innovations in Edinburgh s WMT16 Systems 1 Subword models to allow translation of rare/unknown words since networks have small, fixed vocabulary 2 Back-translated monolingual data as additional training data allows us to make use of extensive monolingual resources 3 Combination of left-to-right and right-to-left models Reduces label-bias problem 4 Pervasive dropout Technical device to improve training with small data sets Barry Haddow Edinburgh s NMT Systems 5 / 20
10 Problem with Word-level Models they charge a carry-on bag fee. sie erheben eine Hand gepäck gebühr. Neural MT architectures have small and fixed vocabulary translation is an open-vocabulary problem productive word formation (example: compounding) names (may require transliteration) Barry Haddow Edinburgh s NMT Systems 6 / 20
11 Why Subword Models? transparent translations many translations are semantically/phonologically transparent translation via subword units possible morphologically complex words (e.g. compounds): solar system (English) Sonnen system (German) Nap rendszer (Hungarian) named entities: Barack Obama (English; German) Áàðàê Îáàìà (Russian) バラク オバマ (ba-ra-ku o-ba-ma) (Japanese) cognates and loanwords: claustrophobia (English) Klaustrophobie (German) Êëàóñòðîôîáèÿ (Russian) Barry Haddow Edinburgh s NMT Systems 7 / 20
12 Byte-pair Encoding Start with maximally split (i.e. characters) Use statistics to identify groups to merge Proceed until pre-determined number of merge operations is learnt Examples: system sentence source health research institutes reference Gesundheitsforschungsinstitute word-level Forschungsinstitute subword Gesundheits forsch ungsin stitute source rakfisk reference ðàêôèñêà (rakfiska) word-level rakfisk UNK rakfisk subword rak f isk ðàê ô èñêà (rak f iska) Barry Haddow Edinburgh s NMT Systems 8 / 20
13 Subword Models translation of many rare/unknown words is transparent subword units allow open-vocabulary NMT without back-off model substantial gains in translation quality, especially for rare words Barry Haddow Edinburgh s NMT Systems 9 / 20
14 Monolingual Training Data why monolingual data for phrase-based SMT? relax independence assumptions more training data more appropriate training data (domain adaptation) why monolingual data for neural MT? relax independence assumptions more training data more appropriate training data (domain adaptation) Barry Haddow Edinburgh s NMT Systems 10 / 20
15 Monolingual Data in NMT encoder-decoder already conditions on previous target words no architecture change required to learn from monolingual data Barry Haddow Edinburgh s NMT Systems 11 / 20
16 Monolingual Training Instances Problem we have no source context for monolingual training instances Barry Haddow Edinburgh s NMT Systems 12 / 20
17 Monolingual Training Instances Problem we have no source context for monolingual training instances Solutions two methods to deal with missing source context: empty/dummy source context danger of unlearning conditioning on source produce synthetic source sentence via back-translation get approximation of source context Barry Haddow Edinburgh s NMT Systems 12 / 20
18 Evaluation: WMT 15 English German BLEU 20 BLEU syntax-based parallel +monolingual +synthetic 0 PBSMT parallel +synthetic +synth-ens4 English German German English Barry Haddow Edinburgh s NMT Systems 13 / 20
19 Why is monolingual data helpful? Domain adaptation effect Reduces over-fitting Improves fluency Barry Haddow Edinburgh s NMT Systems 14 / 20
20 Putting it all together: WMT16 Results BLEU EN CS EN DE EN RO EN RU CS EN DE EN RO EN RU EN parallel data +synthetic data +ensemble +R2L reranking Barry Haddow Edinburgh s NMT Systems 15 / 20
21 Human Pairwise Ranking Scores direction BLEU rank human rank EN CS 1 of 9 1 of 20 EN DE 1 of 11 1 of 15 EN RO 2 of of 12 EN RU 1 of of 12 CS EN 1 of 4 1 of 12 DE EN 1 of 6 1 of 10 RO EN 2 of 5 2 of 7 RU EN 3 of 6 5 of 10 NB: Human rankings include online systems, and for en cs, extra systems from tuning task Barry Haddow Edinburgh s NMT Systems 16 / 20
22 NMT vs. PBMT: An extended test Training and test drawn from UN corpus Multi-parallel, 11M lines Arabic, Chinese, English, French, Russian, Spanish Apply BPE, but no monolingual data NMT systems trained for 8 days Evaluate using BLEU on 4000 sentences [Junczys-Dowmunt et. al, 2016] Barry Haddow Edinburgh s NMT Systems 17 / 20
23 NMT vs. PBMT: UN data Barry Haddow Edinburgh s NMT Systems 18 / 20
24 Conclusions and Outlook Conclusions neural MT is SOTA on many tasks subword models and back-translated data contributed to success NMT has gone from lab to deployment, much faster than expected Problems and Opportunities Inclusion of terminology, placeholders, markup Interpretation and manual changes to models Computer-aided and interactive MT Incremental training Extra knowledge sources (context, multimodal) Sharing across languages, domains Barry Haddow Edinburgh s NMT Systems 19 / 20
25 Acknowledgments Some of the research presented here was conducted in cooperation with Samsung Electronics Polska sp. z o.o. - Samsung R&D Institute Poland. This project has received funding from the European Union s Horizon 2020 research and innovation programme under grant agreement (QT21). Barry Haddow Edinburgh s NMT Systems 20 / 20
26 Thank-you Barry Haddow Edinburgh s NMT Systems 21 / 20
27 WMT16 Results (BLEU) uedin-nmt 34.2 metamind 32.3 NYU-UMontreal 30.8 cambridge 30.6 uedin-syntax 30.6 KIT/LIMSI 29.1 KIT 29.0 uedin-pbmt 28.4 jhu-syntax 26.6 EN DE uedin-nmt 38.6 uedin-pbmt 35.1 jhu-pbmt 34.5 uedin-syntax 34.4 KIT 33.9 jhu-syntax 31.0 DE EN uedin-nmt 25.8 NYU-UMontreal 23.6 jhu-pbmt 23.6 cu-chimera 21.0 uedin-cu-syntax 20.9 cu-tamchyna 20.8 cu-tectomt 14.7 cu-mergedtrees 8.2 EN CS uedin-nmt 31.4 jhu-pbmt 30.4 PJATK 28.3 cu-mergedtrees 13.3 CS EN uedin-pbmt 35.2 uedin-nmt 33.9 uedin-syntax 33.6 jhu-pbmt 32.2 LIMSI 31.0 RO EN QT21-HimL-SysComb 28.9 uedin-nmt 28.1 RWTH-SYSCOMB 27.1 uedin-pbmt 26.8 uedin-lmu-hiero 25.9 KIT 25.8 lmu-cuni 24.3 LIMSI 23.9 jhu-pbmt 23.5 usfd-rescoring 23.1 EN RO uedin-nmt 26.0 amu-uedin 25.3 jhu-pbmt 24.0 LIMSI 23.6 AFRL-MITLL 23.5 NYU-UMontreal 23.1 AFRL-MITLL-verb-annot 20.9 EN RU amu-uedin 29.1 NRC 29.1 uedin-nmt 28.0 AFRL-MITLL 27.6 AFRL-MITLL-contrast 27.0 RU EN Barry Haddow Edinburgh s NMT Systems 22 / 20
28 WMT16 Results (BLEU) uedin-nmt 34.2 metamind 32.3 NYU-UMontreal 30.8 cambridge 30.6 uedin-syntax 30.6 KIT/LIMSI 29.1 KIT 29.0 uedin-pbmt 28.4 jhu-syntax 26.6 EN DE uedin-nmt 38.6 uedin-pbmt 35.1 jhu-pbmt 34.5 uedin-syntax 34.4 KIT 33.9 jhu-syntax 31.0 DE EN uedin-nmt 25.8 NYU-UMontreal 23.6 jhu-pbmt 23.6 cu-chimera 21.0 uedin-cu-syntax 20.9 cu-tamchyna 20.8 cu-tectomt 14.7 cu-mergedtrees 8.2 EN CS uedin-nmt 31.4 jhu-pbmt 30.4 PJATK 28.3 cu-mergedtrees 13.3 CS EN uedin-pbmt 35.2 uedin-nmt 33.9 uedin-syntax 33.6 jhu-pbmt 32.2 LIMSI 31.0 RO EN QT21-HimL-SysComb 28.9 uedin-nmt 28.1 RWTH-SYSCOMB 27.1 uedin-pbmt 26.8 uedin-lmu-hiero 25.9 KIT 25.8 lmu-cuni 24.3 LIMSI 23.9 jhu-pbmt 23.5 usfd-rescoring 23.1 EN RO uedin-nmt 26.0 amu-uedin 25.3 jhu-pbmt 24.0 LIMSI 23.6 AFRL-MITLL 23.5 NYU-UMontreal 23.1 AFRL-MITLL-verb-annot 20.9 EN RU amu-uedin 29.1 NRC 29.1 uedin-nmt 28.0 AFRL-MITLL 27.6 AFRL-MITLL-contrast 27.0 RU EN Edinburgh NMT Barry Haddow Edinburgh s NMT Systems 22 / 20
29 WMT16 Results (BLEU) uedin-nmt 34.2 metamind 32.3 NYU-UMontreal 30.8 cambridge 30.6 uedin-syntax 30.6 KIT/LIMSI 29.1 KIT 29.0 uedin-pbmt 28.4 jhu-syntax 26.6 EN DE uedin-nmt 38.6 uedin-pbmt 35.1 jhu-pbmt 34.5 uedin-syntax 34.4 KIT 33.9 jhu-syntax 31.0 DE EN uedin-nmt 25.8 NYU-UMontreal 23.6 jhu-pbmt 23.6 cu-chimera 21.0 uedin-cu-syntax 20.9 cu-tamchyna 20.8 cu-tectomt 14.7 cu-mergedtrees 8.2 EN CS uedin-nmt 31.4 jhu-pbmt 30.4 PJATK 28.3 cu-mergedtrees 13.3 CS EN uedin-pbmt 35.2 uedin-nmt 33.9 uedin-syntax 33.6 jhu-pbmt 32.2 LIMSI 31.0 RO EN QT21-HimL-SysComb 28.9 uedin-nmt 28.1 RWTH-SYSCOMB 27.1 uedin-pbmt 26.8 uedin-lmu-hiero 25.9 KIT 25.8 lmu-cuni 24.3 LIMSI 23.9 jhu-pbmt 23.5 usfd-rescoring 23.1 EN RO uedin-nmt 26.0 amu-uedin 25.3 jhu-pbmt 24.0 LIMSI 23.6 AFRL-MITLL 23.5 NYU-UMontreal 23.1 AFRL-MITLL-verb-annot 20.9 EN RU amu-uedin 29.1 NRC 29.1 uedin-nmt 28.0 AFRL-MITLL 27.6 AFRL-MITLL-contrast 27.0 RU EN Edinburgh NMT System Combination with Edinburgh NMT Barry Haddow Edinburgh s NMT Systems 22 / 20
30 WMT16: English German # score range system UEDIN-NMT METAMIND UEDIN-SYNTAX NYU-MONTREAL ONLINE-B KIT-LIMSI CAMBRIDGE ONLINE-A PROMT-RULE KIT JHU-SYNTAX JHU-PBMT UEDIN-PBMT ONLINE-F ONLINE-G Barry Haddow Edinburgh s NMT Systems 23 / 20
31 WMT16: English German # score range system UEDIN-NMT METAMIND UEDIN-SYNTAX NYU-MONTREAL ONLINE-B KIT-LIMSI CAMBRIDGE ONLINE-A PROMT-RULE KIT JHU-SYNTAX JHU-PBMT UEDIN-PBMT ONLINE-F ONLINE-G Neural MT Barry Haddow Edinburgh s NMT Systems 23 / 20
32 WMT16: English German # score range system UEDIN-NMT METAMIND UEDIN-SYNTAX NYU-MONTREAL ONLINE-B KIT-LIMSI CAMBRIDGE ONLINE-A PROMT-RULE KIT JHU-SYNTAX JHU-PBMT UEDIN-PBMT ONLINE-F ONLINE-G Neural MT Neural components Barry Haddow Edinburgh s NMT Systems 23 / 20
33 Fluency and Adequacy (German English) de-en Adequacy Fluency UEDIN-NMT ONLINE-A ONLINE-B UEDIN-SYNTAX KIT UEDIN-PBMT JHU-PBMT ONLINE-G ONLINE-F JHU-SYNTAX [Bojar et al. WMT 2016] Barry Haddow Edinburgh s NMT Systems 24 / 20
34 Neural MT and Phrase-based SMT Neural MT Phrase-based SMT translation quality model size training time model interpretability decoding efficiency toolkits (for simplicity) (for maturity) special hardware requirement GPU lots of RAM Barry Haddow Edinburgh s NMT Systems 25 / 20
Smaller, faster, deeper: University of Edinburgh MT submittion to WMT 2017
Smaller, faster, deeper: University of Edinburgh MT submittion to WMT 2017 Rico Sennrich, Alexandra Birch, Anna Currey, Ulrich Germann, Barry Haddow, Kenneth Heafield, Antonio Valerio Miceli Barone, Philip
Deep Architectures for Neural Machine Translation
Deep Architectures for Neural Machine Translation Antonio Valerio Miceli Barone Jindřich Helcl Rico Sennrich Barry Haddow Alexandra Birch School of Informatics, University of Edinburgh Faculty of Mathematics
Translation Quality Assessment: Evaluation and Estimation
Translation Quality Assessment: Evaluation and Estimation Lucia Specia University of Sheffield l.specia@sheffield.ac.uk MTM - Prague, 12 September 2016 Translation Quality Assessment: Evaluation and Estimation
Beam Search Strategies for Neural Machine Translation
Beam Search Strategies for Neural Machine Translation Markus Freitag and Yaser Al-Onaizan IBM T.J. Watson Research Center 1101 Kitchawan Rd, Yorktown Heights, NY 10598 {freitagm,onaizan}@us.ibm.com Abstract
Asynchronous Bidirectional Decoding for Neural Machine Translation
The Thirty-Second AAAI Conference on Artificial Intelligence (AAAI-18) Asynchronous Bidirectional Decoding for Neural Machine Translation Xiangwen Zhang, 1 Jinsong Su, 1 Yue Qin, 1 Yang Liu, 2 Rongrong
When to Finish? Optimal Beam Search for Neural Text Generation (modulo beam size)
When to Finish? Optimal Beam Search for Neural Text Generation (modulo beam size) Liang Huang and Kai Zhao and Mingbo Ma School of Electrical Engineering and Computer Science Oregon State University Corvallis,
arxiv: v1 [cs.cl] 16 Jan 2018
![arxiv: v1 [cs.cl] 16 Jan 2018](/thumbs/93/114087936.jpg "arxiv: v1 [cs.cl] 16 Jan 2018")
Asynchronous Bidirectional Decoding for Neural Machine Translation Xiangwen Zhang 1, Jinsong Su 1, Yue Qin 1, Yang Liu 2, Rongrong Ji 1, Hongi Wang 1 Xiamen University, Xiamen, China 1 Tsinghua University,
Machine Translation Contd
Machine Translation Contd Prof. Sameer Singh CS 295: STATISTICAL NLP WINTER 2017 March 7, 2017 Based on slides from Richard Socher, Chris Manning, Philipp Koehn, and everyone else they copied from. Upcoming
Minimum Risk Training For Neural Machine Translation. Shiqi Shen, Yong Cheng, Zhongjun He, Wei He, Hua Wu, Maosong Sun, and Yang Liu
Minimum Risk Training For Neural Machine Translation Shiqi Shen, Yong Cheng, Zhongjun He, Wei He, Hua Wu, Maosong Sun, and Yang Liu ACL 2016, Berlin, German, August 2016 Machine Translation MT: using computer
An Empirical Study of Adequate Vision Span for Attention-Based Neural Machine Translation
An Empirical Study of Adequate Vision Span for Attention-Based Neural Machine Translation Raphael Shu, Hideki Nakayama shu@nlab.ci.i.u-tokyo.ac.jp, nakayama@ci.i.u-tokyo.ac.jp The University of Tokyo In
Further Meta-Evaluation of Machine Translation
Further Meta-Evaluation of Machine Translation Chris Callison-Burch Johns Hopkins University ccb cs jhu edu Cameron Fordyce camfordyce gmail com Philipp Koehn University of Edinburgh pkoehn inf ed ac uk
Further Meta-Evaluation of Machine Translation
Further Meta-Evaluation of Machine Translation Chris Callison-Burch Johns Hopkins University ccb cs jhu edu Cameron Fordyce camfordyce gmail com Philipp Koehn University of Edinburgh pkoehn inf ed ac uk
Translation Quality Assessment: Evaluation and Estimation
Translation Quality Assessment: Evaluation and Estimation Lucia Specia University of Sheffield l.specia@sheffield.ac.uk 9 September 2013 Translation Quality Assessment: Evaluation and Estimation 1 / 33
Exploring Normalization Techniques for Human Judgments of Machine Translation Adequacy Collected Using Amazon Mechanical Turk
Exploring Normalization Techniques for Human Judgments of Machine Translation Adequacy Collected Using Amazon Mechanical Turk Michael Denkowski and Alon Lavie Language Technologies Institute School of
Edinburgh Research Explorer
Edinburgh Research Explorer Further meta-evaluation of machine translation Citation for published version: Callison-Burch, C, Fordyce, C, Koehn, P, Monz, C & Schroeder, J 2008, Further meta-evaluation
Phrase-level Quality Estimation for Machine Translation
Phrase-level Quality Estimation for Machine Translation Varvara Logacheva, Lucia Specia University of Sheffield December 3, 2015 Varvara Logacheva, Lucia Specia Phrase-level Quality Estimation 1/21 Quality
Sign Language MT. Sara Morrissey
Sign Language MT Sara Morrissey Introduction Overview Irish Sign Language Problems for SLMT SLMT Data MaTrEx for SLs Future work Introduction (1) Motivation SLs are poorly resourced and lack political,
Toward Univeral Network-based Speech Translation
Toward Univeral Network-based Speech Translation Chai Wutiwiwatchai Speech and Audio Technology Laboratory National Electronics and Computer Technology Center 1 Outline Technology Review U-STAR Consortium
Corpus Construction and Semantic Analysis of Indonesian Image Description
Corpus Construction and Semantic Analysis of Indonesian Image Description Khumaisa Nur aini 1,3, Johanes Effendi 1, Sakriani Sakti 1,2, Mirna Adriani 3, Sathosi Nakamura 1,2 1 Nara Institute of Science
Toward the Evaluation of Machine Translation Using Patent Information
Toward the Evaluation of Machine Translation Using Patent Information Atsushi Fujii Graduate School of Library, Information and Media Studies University of Tsukuba Mikio Yamamoto Graduate School of Systems
Exploiting Patent Information for the Evaluation of Machine Translation
Exploiting Patent Information for the Evaluation of Machine Translation Atsushi Fujii University of Tsukuba Masao Utiyama National Institute of Information and Communications Technology Mikio Yamamoto
Overview of the Patent Translation Task at the NTCIR-7 Workshop
Overview of the Patent Translation Task at the NTCIR-7 Workshop Atsushi Fujii, Masao Utiyama, Mikio Yamamoto, Takehito Utsuro University of Tsukuba National Institute of Information and Communications
24-7 Emergency line service (English)
Emergency Line 24-7 Emergency line service (English) 1. Incident occurs 2. Caller dials 24-7 Specialist Advice Number. Incident Advisors hear whisper from automated phone system and are connected to caller
Better Hypothesis Testing for Statistical Machine Translation: Controlling for Optimizer Instability
Better Hypothesis Testing for Statistical Machine Translation: Controlling for Optimizer Instability Jonathan H. Clark Chris Dyer Alon Lavie Noah A. Smith Language Technologies Institute Carnegie Mellon
Does enjoying a movie stimulate the incidental acquisition of language? Géry d Ydewalle. University of Leuven Royal Academy of Sciences BELGIUM
Does enjoying a movie stimulate the incidental acquisition of language? Géry d Ydewalle University of Leuven Royal Academy of Sciences BELGIUM Some facts from Belgium..! Trilingual country: 60% Dutch (Flemish),
Measurement of Progress in Machine Translation
Measurement of Progress in Machine Translation Yvette Graham Timothy Baldwin Aaron Harwood Alistair Moffat Justin Zobel Department of Computing and Information Systems The University of Melbourne {ygraham,tbaldwin,aharwood,amoffat,jzobel}@unimelb.edu.au
Unsupervised Measurement of Translation Quality Using Multi-engine, Bi-directional Translation
Unsupervised Measurement of Translation Quality Using Multi-engine, Bi-directional Translation Menno van Zaanen and Simon Zwarts Division of Information and Communication Sciences Department of Computing
Analyzing Optimization for Statistical Machine Translation: MERT Learns Verbosity, PRO Learns Length
Analyzing Optimization for Statistical Machine Translation: MERT Learns Verbosity, PRO Learns Length Francisco Guzmán Preslav Nakov and Stephan Vogel ALT Research Group Qatar Computing Research Institute,
Choosing the Right Evaluation for Machine Translation: an Examination of Annotator and Automatic Metric Performance on Human Judgment Tasks
Choosing the Right Evaluation for Machine Translation: an Examination of Annotator and Automatic Metric Performance on Human Judgment Tasks Michael Denkowski and Alon Lavie Language Technologies Institute
Speech Processing / Speech Translation Case study: Transtac Details
Speech Processing 11-492/18-492 Speech Translation Case study: Transtac Details Phraselator: One Way Translation Commercial System VoxTec Rapid deployment Modules of 500ish utts Transtac: Two S2S System
arxiv: v2 [cs.cl] 4 Sep 2018
![arxiv: v2 [cs.cl] 4 Sep 2018](/thumbs/91/105780802.jpg "arxiv: v2 [cs.cl] 4 Sep 2018")
Training Deeper Neural Machine Translation Models with Transparent Attention Ankur Bapna Mia Xu Chen Orhan Firat Yuan Cao ankurbpn,miachen,orhanf,yuancao@google.com Google AI Yonghui Wu arxiv:1808.07561v2
. Semi-automatic WordNet Linking using Word Embeddings. Kevin Patel, Diptesh Kanojia and Pushpak Bhattacharyya Presented by: Ritesh Panjwani
Semi-automatic WordNet Linking using Word Embeddings Kevin Patel, Diptesh Kanojia and Pushpak Bhattacharyya Presented by: Ritesh Panjwani January 11, 2018 Kevin Patel WordNet Linking via Embeddings 1/22
Neural Machine Translation with Key-Value Memory-Augmented Attention
Neural Machine Translation with Key-Value Memory-Augmented Attention Fandong Meng, Zhaopeng Tu, Yong Cheng, Haiyang Wu, Junjie Zhai, Yuekui Yang, Di Wang Tencent AI Lab {fandongmeng,zptu,yongcheng,gavinwu,jasonzhai,yuekuiyang,diwang}@tencent.com
School Curriculum and Standards Authority
Accounting and Finance Stage 2-74 42 12 57 9 43 21-50 Accounting and Finance Stage 3-1366 1181 596 48 635 52 1231 4.2 Ancient History Stage 2-4 10 2 40 3 60 5-50 Ancient History Stage 3-200 212 92 60 62
Translation User Manual. Don t Translate it. Lingmo it! Leading edge language translation technology for the global market
Translation User Manual Don t Translate it. Lingmo it! Leading edge language translation technology for the global market Contents About T2T Translation Software Your Time2Translate Charging your Time2Translate
Medical Knowledge Attention Enhanced Neural Model. for Named Entity Recognition in Chinese EMR
Medical Knowledge Attention Enhanced Neural Model for Named Entity Recognition in Chinese EMR Zhichang Zhang, Yu Zhang, Tong Zhou College of Computer Science and Engineering, Northwest Normal University,
Image Captioning using Reinforcement Learning. Presentation by: Samarth Gupta
Image Captioning using Reinforcement Learning Presentation by: Samarth Gupta 1 Introduction Summary Supervised Models Image captioning as RL problem Actor Critic Architecture Policy Gradient architecture
Sequential Predictions Recurrent Neural Networks
CS 2770: Computer Vision Sequential Predictions Recurrent Neural Networks Prof. Adriana Kovashka University of Pittsburgh March 28, 2017 One Motivation: Descriptive Text for Images It was an arresting
Appendix I Teaching outcomes of the degree programme (art. 1.3)
Appendix I Teaching outcomes of the degree programme (art. 1.3) The Master graduate in Computing Science is fully acquainted with the basic terms and techniques used in Computing Science, and is familiar
Context Gates for Neural Machine Translation
Context Gates for Neural Machine Translation Zhaopeng Tu Yang Liu Zhengdong Lu Xiaohua Liu Hang Li Noah s Ark Lab, Huawei Technologies, Hong Kong {tu.zhaopeng,lu.zhengdong,liuxiaohua3,hangli.hl}@huawei.com
Introduction to Computational Neuroscience
Introduction to Computational Neuroscience Lecture 5: Data analysis II Lesson Title 1 Introduction 2 Structure and Function of the NS 3 Windows to the Brain 4 Data analysis 5 Data analysis II 6 Single
A Simple Pipeline Application for Identifying and Negating SNOMED CT in Free Text
A Simple Pipeline Application for Identifying and Negating SNOMED CT in Free Text Anthony Nguyen 1, Michael Lawley 1, David Hansen 1, Shoni Colquist 2 1 The Australian e-health Research Centre, CSIRO ICT
Recurrent Neural Networks
CS 2750: Machine Learning Recurrent Neural Networks Prof. Adriana Kovashka University of Pittsburgh March 14, 2017 One Motivation: Descriptive Text for Images It was an arresting face, pointed of chin,
Distillation of Knowledge from the Research Literatures on Alzheimer s Dementia
JSCI 2017 1 Distillation of Knowledge from the Research Literatures on Alzheimer s Dementia Wutthipong Kongburan, Mark Chignell, and Jonathan H. Chan School of Information Technology King Mongkut's University
Deep Learning based Information Extraction Framework on Chinese Electronic Health Records
Deep Learning based Information Extraction Framework on Chinese Electronic Health Records Bing Tian Yong Zhang Kaixin Liu Chunxiao Xing RIIT, Beijing National Research Center for Information Science and
Has Machine Translation Achieved Human Parity? A Case for Document-level Evaluation
Has Machine Translation Achieved Human Parity? A Case for Document-level Evaluation Samuel Läubli 1 Rico Sennrich 1,2 Martin Volk 1 1 Institute of Computational Linguistics, University of Zurich {laeubli,volk}@cl.uzh.ch
Audiovisual to Sign Language Translator
Technical Disclosure Commons Defensive Publications Series July 17, 2018 Audiovisual to Sign Language Translator Manikandan Gopalakrishnan Follow this and additional works at: https://www.tdcommons.org/dpubs_series
English to ASL Gloss Machine Translation
Brigham Young University BYU ScholarsArchive All Theses and Dissertations 2015-06-01 English to ASL Gloss Machine Translation Mary Elizabeth Bonham Brigham Young University - Provo Follow this and additional
Language Services / Tasks
Athena Kartalou Language Services Manager Games Services Special Olympics ATHENS 2011 Language Services / Tasks To provide linguistic services to delegations coming from all over the world. To cover as
Memory-Augmented Active Deep Learning for Identifying Relations Between Distant Medical Concepts in Electroencephalography Reports
Memory-Augmented Active Deep Learning for Identifying Relations Between Distant Medical Concepts in Electroencephalography Reports Ramon Maldonado, BS, Travis Goodwin, PhD Sanda M. Harabagiu, PhD The University
Text mining for lung cancer cases over large patient admission data. David Martinez, Lawrence Cavedon, Zaf Alam, Christopher Bain, Karin Verspoor
Text mining for lung cancer cases over large patient admission data David Martinez, Lawrence Cavedon, Zaf Alam, Christopher Bain, Karin Verspoor Opportunities for Biomedical Informatics Increasing roll-out
Adjudicator Agreement and System Rankings for Person Name Search
Adjudicator Agreement and System Rankings for Person Name Search Mark D. Arehart, Chris Wolf, Keith J. Miller The MITRE Corporation 7515 Colshire Dr., McLean, VA 22102 {marehart, cwolf, keith}@mitre.org
Convolutional and LSTM Neural Networks
Convolutional and LSTM Neural Networks Vanessa Jurtz January 11, 2017 Contents Neural networks and GPUs Lasagne Peptide binding to MHC class II molecules Convolutional Neural Networks (CNN) Recurrent and
Summary. About Pilot Waverly Labs is... Core Team Pilot Translation Kit FAQ Pilot Speech Translation App FAQ Testimonials Contact Useful Links. p.
Press kit Summary About Pilot Waverly Labs is... Core Team Pilot Translation Kit FAQ Pilot Speech Translation App FAQ Testimonials Contact Useful Links p. 3 p.4 p. 5 p. 6 p. 8 p. 10 p. 12 p. 14 p. 16 p.
Chapter 12 Conclusions and Outlook
Chapter 12 Conclusions and Outlook In this book research in clinical text mining from the early days in 1970 up to now (2017) has been compiled. This book provided information on paper based patient record
WAVERLY LABS - Press Kit
Press kit Summary About Pilot Waverly Labs Core Team Pilot Translating Earpiece FAQ Pilot Speech Translator App FAQ Testimonials Contact Useful Links p. 3 p.4 p. 5 p. 6 p. 8 p. 10 p. 12 p. 14 p. 16 p.
What Predicts Media Coverage of Health Science Articles?
What Predicts Media Coverage of Health Science Articles? Byron C Wallace, Michael J Paul & Noémie Elhadad University of Texas at Austin; byron.wallace@utexas.edu! Johns Hopkins University! Columbia University!
Efficient Attention using a Fixed-Size Memory Representation
Efficient Attention using a Fixed-Size Memory Representation Denny Britz and Melody Y. Guan and Minh-Thang Luong Google Brain dennybritz,melodyguan,thangluong@google.com Abstract The standard content-based
Normalising orthographic and dialectal variants for the automatic processing of Swiss German Tanja Samardžić, Yves Scherrer, Elvira Glaser
Normalising orthographic and dialectal variants for the automatic processing of Swiss German Tanja Samardžić, Yves Scherrer, Elvira Glaser 27.12.2015 Page 1 Swiss German Dialects Map by Péter Jeszenszky
GSE Teacher Toolkit Download
Teacher Toolkit Download Thanks for using the Teacher Toolkit - here are your search results. Have you tried using the Toolkit to audit your school's curriculum? Or to help plan your lessons? Find out
Convolutional and LSTM Neural Networks
Convolutional and LSTM Neural Networks Vanessa Jurtz January 12, 2016 Contents Neural networks and GPUs Lasagne Peptide binding to MHC class II molecules Convolutional Neural Networks (CNN) Recurrent and
#5411 Educational Leadership: Administration and Supervision. #5135: Art: Content & Analysis Knowledge. #5101 Business Education: Content Knowledge
Colorado Endorsement Praxis Exam Passing Score Administrator (K- 145 Agriculture and Renewable Natural Resources (7- Art (K- #5701 Agriculture 147 #5135: Art: Content & Analysis 158 American Sign Language
Massive Exploration of Neural Machine Translation Architectures
Massive Exploration of Neural Machine Translation Architectures Denny Britz, Anna Goldie, Minh-Thang Luong, Quoc V. Le {dennybritz,agoldie,thangluong,qvl}@google.com Google Brain Abstract Neural Machine
NILab. NeuroInformatics Laboratory
NILab NeuroInformatics Laboratory NILab Premise Office Building, Mattarello Center for Information Technology Fondazione Bruno Kessler www.fbk.eu Center for Mind/Brain Sciences Università degli Studi di
Overview of the visual cortex. Ventral pathway. Overview of the visual cortex
Overview of the visual cortex Two streams: Ventral What : V1,V2, V4, IT, form recognition and object representation Dorsal Where : V1,V2, MT, MST, LIP, VIP, 7a: motion, location, control of eyes and arms
Continuous Sign Language Recognition Approaches from Speech Recognition
Continuous Sign Language Recognition Approaches from Speech Recognition Philippe Dreuw dreuw@i6.informatik.rwth-aachen.de Invited Talk at DCU 9. June 2006 Human Language Technology and Pattern Recognition
Incorporating Word Reordering Knowledge into. attention-based Neural Machine Translation
Incorporating Word Reordering Knowledge into Attention-based Neural Machine Translation Jinchao Zhang 1 Mingxuan Wang 1 Qun Liu 3,1 Jie Zhou 2 1 Key Laboratory of Intelligent Information Processing, Institute
Annotation and Retrieval System Using Confabulation Model for ImageCLEF2011 Photo Annotation
Annotation and Retrieval System Using Confabulation Model for ImageCLEF2011 Photo Annotation Ryo Izawa, Naoki Motohashi, and Tomohiro Takagi Department of Computer Science Meiji University 1-1-1 Higashimita,
TWO HANDED SIGN LANGUAGE RECOGNITION SYSTEM USING IMAGE PROCESSING
134 TWO HANDED SIGN LANGUAGE RECOGNITION SYSTEM USING IMAGE PROCESSING H.F.S.M.Fonseka 1, J.T.Jonathan 2, P.Sabeshan 3 and M.B.Dissanayaka 4 1 Department of Electrical And Electronic Engineering, Faculty
Course Introduction. Neural Information Processing: Introduction. Notes. Administration
3 / 17 4 / 17 Course Introduction Neural Information Processing: Introduction Matthias Hennig and Mark van Rossum School of Informatics, University of Edinburgh Welcome and administration Course outline
UNGASS 2016 Global Civil Society Survey
UNGASS 2016 Global Civil Society Survey Preliminary Results and Report S H E I L A P. VA K H A R I A P H. D., L. M. S.W. LO N G I S L A ND U N I V ERS I T Y B R O OKLY N, N E W YO R K, U S A L I NDA N
Summer 2016 Examinations Timetable. Pm exams start 13:30 + Period Day / Date Period 1 Period 2 Period 3 Period 4.
02/05/16 03/05/16 English as First Language reading - 1hr 45m (12) 04/05/16 05/05/16 06/05/16 English as First Language writing 2 hours (12) 09/05/16 10/05/16 11/05/16 12/05/16 13/05/16 Welsh as Second
INTERLINGUAL SUBTITLES AND SDH IN HBBTV
INTERLINGUAL SUBTITLES AND SDH IN HBBTV CONTENTS 0. Definition 2 1. DisTINCTIVE FEATURES OF SDH 3 2. PRESENTATION 5 3. DISPLAY ON THE SCREEN 6 4. FONT SIZE 7 5. FONT TYPE 8 6. COLOUR AND CONTRAST 8 7.
Suicide prevention from a global perspective Dr Alexandra Fleischmann Department of Mental Health and Substance Abuse World Health Organization
San José, Costa Rica, May 2016 Suicide prevention from a global perspective Dr Alexandra Fleischmann Department of Mental Health and Substance Abuse World Health Organization Suicide facts (1) Over 800
Information for Candidates Preparing for the CILISAT
Information for Candidates Preparing for the CILISAT CILISAT Community Interpreter Language and Interpreting Skills Assessment Tool Cultural Interpretation Services for Our Communities 44 Eccles Street,
arxiv: v1 [cs.cl] 17 Oct 2016
![arxiv: v1 [cs.cl] 17 Oct 2016](/thumbs/85/91747772.jpg "arxiv: v1 [cs.cl] 17 Oct 2016")
Interactive Attention for Neural Machine Translation Fandong Meng 1 Zhengdong Lu 2 Hang Li 2 Qun Liu 3,4 arxiv:1610.05011v1 [cs.cl] 17 Oct 2016 1 AI Platform Department, Tencent Technology Co., Ltd. fandongmeng@tencent.com
DATA is derived either through. Self-Report Observation Measurement
Data Management DATA is derived either through Self-Report Observation Measurement QUESTION ANSWER DATA DATA may be from Structured or Unstructured questions? Quantitative or Qualitative? Numerical or
Modeling the Use of Space for Pointing in American Sign Language Animation
Modeling the Use of Space for Pointing in American Sign Language Animation Jigar Gohel, Sedeeq Al-khazraji, Matt Huenerfauth Rochester Institute of Technology, Golisano College of Computing and Information
Processing of Logical Functions in the Human Brain
Processing of Logical Functions in the Human Brain GRMELA ALEŠ AGCES Ltd. AGCES, Levského 3221/1, Praha 4 CZECH REPUBLIC N. E. MASTORAKIS Military Institutions of University Education, Hellenic Naval Academy,
POLICY and PROCEDURE
POLICY and PROCEDURE Interpreter Services Policy Number: ADM-1014 Administration Manual: Administration Reviewed/Revised: 9/21/2016 Effective: I. PURPOSE: To define Martin Luther King, Jr. Community Hospital
Foundations of Natural Language Processing Lecture 13 Heads, Dependency parsing
Foundations of Natural Language Processing Lecture 13 Heads, Dependency parsing Alex Lascarides (slides from Alex Lascarides, Henry Thompson, Nathan Schneider and Sharon Goldwater) 6 March 2018 Alex Lascarides
week by week in the Training Calendar. Follow up graphics, sort, filter, merge, and split functions are also included.
Firstbeat ATHLETE The story Firstbeat ATHLETE is one of the most advanced software tools for analyzing heart rate based training. It makes professional level training analysis available to all dedicated
Language Technologies and Business in the Future
Language Technologies and Business in the Future Niko Papula, Multilizer KITES Symposium, Helsinki 31.10.2013 Finnish company providing leading translation & localization technologies since 1995 Each month
Address Tel: Fax:
Address: Athena Court, Office 32, 3 rd floor, 2 Americanas Street,Potamos Yermasoyias, 4048 Limassol- Cyprus P.O.Box 51673 CY-3507 Limasol - Cyprus Tel: +357 25 325544 Fax: +357 25 318885 Email: info@interbrics.com
20 June 2018 Our Ref: RFI 23569
Information Governance 20 June 2018 Our Ref: RFI 23569 Dear Freedom of Information Act 2000 Information in Relation to Translator Costs I am writing to confirm that the South Eastern Health & Social Care
Interact-AS. Use handwriting, typing and/or speech input. The most recently spoken phrase is shown in the top box
Interact-AS One of the Many Communications Products from Auditory Sciences Use handwriting, typing and/or speech input The most recently spoken phrase is shown in the top box Use the Control Box to Turn
UKParl: A Data Set for Topic Detection with Semantically Annotated Text
UKParl: A Data Set for Topic Detection with Semantically Annotated Text Federico Nanni, Mahmoud Osman, Yi-Ru Cheng, Simone Paolo Ponzetto and Laura Dietz My Research Post-Doc in computational social science
Classification of Epileptic Seizure Predictors in EEG
Classification of Epileptic Seizure Predictors in EEG Problem: Epileptic seizures are still not fully understood in medicine. This is because there is a wide range of potential causes of epilepsy which
Increasing Access to Technical Science Vocabulary Through Use of Universally Designed Signing Dictionaries
UNIVERSAL DESIGN in Higher Education P R O M I S I N G P R A C T I C E S Increasing Access to Technical Science Vocabulary Through Use of Universally Designed Signing Dictionaries Judy Vesel and Tara Robillard,
News English.com Ready-to-Use English Lessons by Sean Banville Level 6 Euro politicians told to stop speaking English
www.breaking News English.com Ready-to-Use English Lessons by Sean Banville 1,000 IDEAS & ACTIVITIES FOR LANGUAGE TEACHERS www.breakingnewsenglish.com/book.html Thousands more free lessons from Sean's
Neural Information Processing: Introduction
1 / 17 Neural Information Processing: Introduction Matthias Hennig School of Informatics, University of Edinburgh January 2017 2 / 17 Course Introduction Welcome and administration Course outline and context
Why did the network make this prediction?
Why did the network make this prediction? Ankur Taly (Google Inc.) Joint work with Mukund Sundararajan and Qiqi Yan Some Deep Learning Successes Source: https://www.cbsinsights.com Deep Neural Networks
DANGEROUS GOODS PANEL (DGP)
International Civil Aviation Organization WORKING PAPER 24/10/05 English only DANGEROUS GOODS PANEL (DGP) TWENTIETH MEETING Montréal, 24 October to 4 November 2005 Agenda Item 2: Development of recommendations
GfK Verein. Detecting Emotions from Voice
GfK Verein Detecting Emotions from Voice Respondents willingness to complete questionnaires declines But it doesn t necessarily mean that consumers have nothing to say about products or brands: GfK Verein
Job (4) Job group Job family Short description Typical CO grade
Job (4) Job group Job family Short description Typical CO grade Scientist Core business Scientists Scientist dealing with seasonal forecast production in a research institute or in a private company producing
Biomedical resources for text mining
August 30, 2005 Workshop Terminologies and ontologies in biomedicine: Can text mining help? Biomedical resources for text mining Olivier Bodenreider Lister Hill National Center for Biomedical Communications
Extraction of Adverse Drug Effects from Clinical Records
MEDINFO 2010 C. Safran et al. (Eds.) IOS Press, 2010 2010 IMIA and SAHIA. All rights reserved. doi:10.3233/978-1-60750-588-4-739 739 Extraction of Adverse Drug Effects from Clinical Records Eiji Aramaki
Convolutional Neural Networks for Text Classification
Convolutional Neural Networks for Text Classification Sebastian Sierra MindLab Research Group July 1, 2016 ebastian Sierra (MindLab Research Group) NLP Summer Class July 1, 2016 1 / 32 Outline 1 What is