IRSTLM (IRST Language Modeling) Toolkit 2014

Notes:

IRSTLM is a tool for the estimation, representation, and computation of statistical language models.

Resources:

  • rnnlm.org .. Recurrent Neural Network Language Models
  • statmt.org/ngrams .. N-gram counts and language models from the CommonCrawl

Wikipedia:

See also:

IRSTLM (IRST Language Modeling) Toolkit 2013


Assamese-English Bilingual Machine Translation KK Baruah, P Das, A Hannan, SK Sarma – arXiv preprint arXiv:1407.2019, 2014 – arxiv.org … To develop the language model and to align the words we used two another tools IRSTLM, GIZA respectively. … Other translation tools like IRSTLM are used to develop language model and GIZA++ to align the words of source text to target one. …

MT systems for the first development cycle AT DCU, S Cortés-Vaíllo, S Ortiz-Rojas… – abumatran.eu … systems, we have used MGIZA (Gao and Vogel., 2008) 0.7.31 to align the corpora, Moses (Koehn et al., 2007) 1.02 to train a regular (non-hierarchical) phrase-based statistical machine translation system with a maximum phrase length of 7 words, and IRSTLM (Federico et al …

Postech’s system description for medical text translation task J Li, SJ Kim, H Na, JH Lee – ACL 2014, 2014 – aclweb.org … For each monolingual corpus, we used a five- gram language model, which was built by IRSTLM toolkit2 (Federico, Bertoldi, & Cettolo, 2008) with improved Kneser Ney smoothing (Chen & Goodman, 1996; Kneser & Ney, 1995). … net/projects/irstlm 229 Page 250. …

Unsupervised POS Tagging of Low Resource Language for Machine Translation YK Thu, A Tamura, A Finch, E Sumita, Y Sagisaka – anlp.jp … Table 1 shows that “Both (POS)” for my- en translation outperforms the baseline by a large margin (+2.40 BLEU). In Table 2, “Both 6http://sourceforge.net/apps/mediawiki/ irstlm/index.php?title= IRSTLM (POS)” achieved a higher RIBES for en-my than the baseline. … Related articles

Comparison of Rule-based and Statistical Methods for Grapheme to Phoneme Modelling I Auzi?a, M Pinnis, R DAR?IS – Human Language Technologies– …, 2014 – researchgate.net … Language models for the SMT systems were built using IRSTLM [5] on the phonetic transcriptions. … [5] Federico, M., Bertoldi, N., & Cettolo, M. IRSTLM: an open source toolkit for handling large scale language models. In Interspeech, (2008), 1618-1621. … Cited by 1

Subsegmental language detection in Celtic language text FM Tyers, A Minocha – 2014 – munin.uit.no … We first built character language models using IRSTLM (Federico et al., 2008) for the five languages in question. For English and French a model was trained using the EuroParl corpus (Koehn, 2005). … Irstlm: an open source toolkit for handling large scale language models. …

Subsegmental language detection in Celtic language text A Minocha, FM Tyers – CLTW 2014, 2014 – aclweb.org … We first built character language models using IRSTLM (Federico et al., 2008) for the five languages in question. For English and French a model was trained using the EuroParl corpus (Koehn, 2005). … Irstlm: an open source toolkit for handling large scale language models. …

SMT for restricted sublanguage in CAT tool context at the European Parliament N Hajlaoui – Conference Chairs and Editors of the Proceedings – tradulex.com … The language models for French were 3-gram ones over each training domain data using the IRSTLM toolkit (Federico, 2008). We used Page 231. … Federico, MN (2008). IRSTLM: an open source toolkit for handling large scale language models. In Proceedings of Interspeech. …

TECHLIMED System Description for the Shared Task on Automatic Arabic Error Correction D MOSTEFA, O ASBAYOU, R ABBES – ANLP 2014, 2014 – aclweb.org … The language model was created with the IRSTLM toolkit (Federico, 2008). We evaluated the translation models on the de- velopment set using different sizes of monolin- gual corpus. … 2008. Irstlm: an open source toolkit for han- dling large scale language models. … Cited by 1

Direct Word Graph Rescoring Using A* Search and RNNLM S Jalalvand, D Falavigna – Fifteenth Annual Conference of the …, 2014 – 193.6.4.39 … news” according to increasing perplexity (computed with the previously mentioned in-domain LM) and we selected the documents with lowest perplexities to build a corpus of about 100 million words [15], over which a 4-gram back-off LM was trained with the IRSTLM toolkit [16 …

Experiments in medical translation shared task at wmt 2014 J Zhang, X Wu, I Calixto, AH Vahid, X Zhang… – ACL …, 2014 – anthology.aclweb.org … The language model was created with open source IRSTLM toolkit (Federico et al., 2008) using all the English in-domain data (monolingual and par- allel). … 2008. IRSTLM: an open source toolkit for handling large scale language models, Interspeech, Brisbane, Australia. …

Statistical Machine Translation H Hassan, K Darwish – Natural Language Processing of Semitic …, 2014 – Springer … Arabic statistical machine translation. Mach Transl 26:25–45 CrossRef; Federico M, Bertoldi N, Cettolo M (2008) IRSTLM: an open source toolkit for handling large scale language models. In: Proceedings of interspeech, Brisbane; … Related articles

Active Learning for Post-Editing Based Incrementally Retrained MT ADJ van Genabith Qun, LJJA Toral – EACL 2014, 2014 – aclweb.org … The statistics of the data (training and in- cremental splits) are shown in Table 1. All the systems are trained using the Moses (Koehn et al., 2007) phrase-based sta- tistical MT system, with IRSTLM (Federico et al., 2008) for language modelling (n-grams up to order five) and with … Related articles All 2 versions

Language model adaptation for automatic call transcription A Haznedaroglu, LM Arslan – Acoustics, Speech and Signal …, 2014 – ieeexplore.ieee.org … LM training and adaptations are done using the IRSTLM toolkit [18]. … [18] M. Federico, N. Bertoldi, M. Cettolo, IRSTLM: an Open Source Toolkit for Handling Large Scale Language Models, Proceedings of Interspeech, Brisbane, Australia, 2008. 4106

UM-Corpus: a large English-Chinese parallel corpus for statistical machine translation L Tian, D Wong, L Chao, P Quaresma… – Proceedings of the 9th …, 2014 – lrec-conf.org … The language model is created by the external toolkits IRSTLM (Federico, 2008) from all the UM-Corpus with 5-gram model. … Federico, M., Bertoldi, N., Cettolo, M. (2008). IRSTLM: an Open Source Toolkit for Handling Large Scale Language Models. … Cited by 5 Related articles

Query-based composition for large-scale language model in LVCSR Y Han, C Zhang, X Li, Y Liu… – Acoustics, Speech and …, 2014 – ieeexplore.ieee.org … and apply them in the systems. Moreover, for language models, some tools such as KenLM[7], SRILM[8], IRSTLM[9] all bring the con- venience to access large-scale language models. As language models are getting larger …

Bootstrapping of a Multilingual Transliteration Dictionary for European Languages M Pinnis – researchgate.net … New pairs from 5th iteration Page 15. AUTOMATIC EVALUATION SCENARIO ? Character-based SMT transliteration task ? IRSTLM language models ? 5-gram language models ? built on all words contained in the initial dictionaries ? Translation models ? 7-gram phrase tables …

N-gram counts and language models from the common crawl C Buck, K Heafield, B van Ooyen – 2014 – lrec-conf.org … As a result, all models sum to 1 over the entire vocabulary. We empha- size that this approach is not new, but rather standard prac- tice recommended by IRSTLM (Federico et al., 2008). … 2008. IRSTLM: an open source toolkit for handling large scale language models. … Cited by 8 Related articles All 2 versions

caWaC–A web corpus of Catalan and its application to language modeling and machine translation N Ljubešic, A Toral – lrec-conf.org … All the LMs are built with the IRSTLM toolkit (Federico et al., 2008), they consider n-grams up to order 5 and they are smoothed using a simplified version of the modified Kneser-Ney method (Chen and Goodman, 1996). The LMs are evaluated for both tasks on four test sets. … Cited by 1 Related articles

Grammatical error correction using hybrid systems and type filtering M Felice, Z Yuan, ØE Andersen, H Yannakoudakis… – CoNLL- …, 2014 – comp.nus.edu.sg … 2014). The IRSTLM Toolkit (Federico et al., 2008) was used to build a 4-gram target language model with Kneser–Ney smoothing (Kneser and Ney, 1995) on the correct sentences from the NUCLE, full CLC and EVP corpora. … Cited by 2 Related articles

Syntactic SMT Using a Discriminative Text Generation Model Y Zhang, K Song, L Song, J Zhu, Q Liu – aclweb.org … 3http://www.nlplab.com/NiuPlan/NiuTrans.ch.html 4http://sourceforge.net/projects/zpar/ 5http://sourceforge.net/apps/mediawiki/irstlm System T2S S2T T2T OURS BLEU 32.65 36.07 28.46 34.24 … 2008. IRSTLM: an open source toolkit for handling large scale language models. …

Language Identification in Code-Switching Scenario N Jain, IH LTRC, RA Bhat – EMNLP 2014, 2014 – aclweb.org … is es- timated from the respective training sets shown in Table 2. Each training set is used to train a separate letter-based language model to estimate the probability of word w. The language model p (w) is implemented as an n-gram model using the IRSTLM-Toolkit (Federico et … Cited by 1

Automatic Speech Recognition and Translation of a Swiss German Dialect: Walliserdeutsch PN Garner, D Imseng, T Meyer – Proceedings of Interspeech, 2014 – infoscience.epfl.ch … However, if the performance is not good, it can be attributed to data sparsity. 4.2. SMT models All DE data was then used to build a language model with the IRSTLM toolkit [13] as an additional feature component of the SMT system. …

A language-independent and fully unsupervised approach to lexicon induction and part-of-speech tagging for closely related languages Y Scherrer, B Sagot – lrec-conf.org … Our C-SMT model relies on the standard pipeline consisting of GIZA++ (Och and Ney, 2003) for character alignment, IRSTLM (Federico et al., 2008) for language modelling, and Moses (Koehn et al., 2007) for phrase extraction and decoding. … Cited by 1 Related articles

Comparing the Quality of Focused Crawlers and of the Translation Resources Obtained from them B Laranjeira, VP Moreira, A Villavicencio, C Ramisch… – 2014 – lrec-conf.org … The resulting sentences were used as input to IRSTLM (Federico et al., 2008), which generated a trigram lan- guage model for the corpus collected by each of the fo- … Irstlm: an open source toolkit for handling large scale language models. In Interspeech, pages 1618–1621. … Cited by 2 Related articles

Statistical Machine Translation N Durrani – 2014 – statmt.org … Moses (Koehn et. al 2007), Phrasal (Cerr et. al 2010), NCode (Crego et. al 2011) – GIZA++ (Word Alignments) – SRILM, IRSTLM, KENLM, LMPLZ (Language Model) • Data – French-English 39M – Chinese-English Spanish-English, Czech-English 15M – Arabic-English …

Adapting Predicate Frames for Urdu PropBanking RA Bhat, N Jain, DM Sharma, A Vaidya, M Palmer… – LT4CloseLang …, 2014 – alt.qcri.org … is esti- mated from the respective training sets shown in Table (3). Each training set is used to train a separate letter-based language model to estimate the probability of word w. The language model p (w) is implemented as an n-gram model using the IRSTLM-Toolkit (Federico …

Optimized MT Online Learning in Computer Assisted Translation P Mathur, M Cettolo – Proceedings of AMTA Workshop on Interactive and …, 2014 – hlt.fbk.eu … 5-gram language models for each task, smoothed through the improved Kneser-Ney technique (Chen and Goodman, 1998), are estimated by means of the IRSTLM toolkit (Federico et al., 2008) on the target side of the training parallel corpora. … Cited by 1

Linguistically-augmented perplexity-based data selection for language models A Toral, P Pecina, L Wang, J van Genabith – Computer Speech & Language, 2014 – Elsevier … All the LMs used in the experiments are built with IRSTLM 5.80.01 (Federico et al., 2008), they consider n-grams up to order 4 and they are smoothed using a simplified version of the modified Kneser-Ney method ( Chen and Goodman, 1996). …

Statistical models for text normalization and machine translation K Scannell – CLTW 2014, 2014 – aclweb.org … discounting (Chen and Goodman, 1996). The implementation of the language model is included as part of the translator itself in order to avoid external dependencies on libraries such as IRSTLM, KenLM, etc. 2.2 Translation Model The …

DCEP-Digital Corpus of the European Parliament N Hajlaoui, D Kolovratnik, J Väyrynen, R Steinberger… – lrec-conf.org Page 1. DCEP -Digital Corpus of the European Parliament Najeh Hajlaoui 1 , David Kolovratnik 1 , Jaakko Väyrynen 2 , Ralf Steinberger 2 , Daniel Varga 3 European Parliament1, European Commission2, Budapest University … Cited by 1 Related articles

A Hybrid Approach to the Development of Bidirectional English-Oromiffa Machine Translation J Daba, Y Assabie – Advances in Natural Language Processing, 2014 – Springer … Thus, we compute p(e) and p(o) as language models for Oromiffa-English and English-Oromiffa translations, respectively. We used n-grams to compute language models with the help of IRSTLM tool. Smoothing was also employed to avoid zero probability for unseen n-grams. …

A Relationship: Word Alignment, Phrase Table, and Translation Quality L Tian, DF Wong, LS Chao, F Oliveira – The Scientific World Journal, 2014 – hindawi.com The Scientific World Journal is a peer-reviewed, open access journal covering a wide range of subjects in science, technology, and medicine. The journal’s Editorial Board as well as its Table of Contents are divided into 98 subject areas that are covered within the journal’s scope … Related articles All 4 versions

Identification of Bilingual Terms from Monolingual Documents for Statistical Machine Translation M Arcan, C Giuliano, M Turchi, P Buitelaar – Proceedings of the 4th …, 2014 – aclweb.org … The IRSTLM toolkit (Federico et al., 2008) was used to build the 5-gram language model. … Marcello Federico, Nicola Bertoldi, and Mauro Cettolo. 2008. Irstlm: an open source toolkit for handling large scale language models. In INTERSPEECH, pages 1618–1621. ISCA. … Cited by 2

Domain and Dialect Adaptation for Machine Translation into Egyptian Arabic S Jeblee, W Feely, H Bouamor, A Lavie, N Habash… – ANLP 2014, 2014 – aclweb.org … using the target side of the training set, and for the core system we used the large MSA language model described in sec- tion 4. We use KenLM because it has been shown (Heafield, 2011) to be faster and use less memory than SRILM (Stolcke, 2002) and IRSTLM (Fed- erico …

Online Multi-User Adaptive Statistical Machine Translation P Mathur, M Cettolo, M Federico, JGC de Souza – matecat.com … 5-gram language models for each task were estimated by means of IRSTLM toolkit (Federico et al., 2008), with improved Kneser-Ney smoothing (Chen and Goodman, 1998), on the target side of the training parallel corpora. …

Quality Estimation for Automatic Speech Recognition M Negri, M Turchi, JGC de Souza, D Falavigna – anthology.aclweb.org … LM probability is computed with a 4-gram backoff LM, trained over about 5 billion words using the IRSTLM toolkit (Federico et al., 2008) and the modified shift-beta smoothing method. … 2008. IRSTLM: an Open Source Toolkit for Handling Large Scale Language Models. …

Generating artificial errors for grammatical error correction M Felice, Z Yuan – EACL 2014, 2014 – aclweb.org … Our setup is similar to the one described by Yuan and Felice (2013) in that we train a PoS-factored phrase-based model (Koehn, 2010) using Moses (Koehn et al., 2007), Giza++ (Och and Ney, 2003) for word alignment and the IRSTLM Toolkit (Federico et al., 2008) for lan … Cited by 1 Related articles All 2 versions

Pivot translation using source-side dictionary and target-side parallel corpus towards MT from resource-limited languages T Nomura, T Akiba – Advanced Informatics: Concept, Theory …, 2014 – ieeexplore.ieee.org … The 5-gram language models were trained using the IRSTLM toolkit [5]. For the conventional pivot-translation method, we applied a cascade of two translation systems proposed by [1]. The method consists of building two machine translation systems: one is the source–pivot …

A Systematic Comparison of Data Selection Criteria for SMT Domain Adaptation L Wang, DF Wong, LS Chao, Y Lu… – The Scientific World …, 2014 – hindawi.com The Scientific World Journal is a peer-reviewed, open access journal covering a wide range of subjects in science, technology, and medicine. The journal’s Editorial Board as well as its Table of Contents are divided into 98 subject areas that are covered within the journal’s scope … Cited by 4 Related articles All 6 versions

Data Selection for Discriminative Training in Statistical Machine Translation X Song, L Specia, T Cohn – people.eng.unimelb.edu.au … The word alignment and 2The 200 sentence pair limit is used to reduce the runtime on large datasets. language models were learned using GIZA++ and IRSTLM with Moses default settings. A trigram language model was trained on English side of the parallel data. … Related articles

Data Selection for Compact Adapted SMT Models S Mirkin, L Besacier – researchgate.net … Table 3: Bilingual corpora for training. SMT system and preprocessing We used Moses (Koehn et al., 2007) as our phrase-based SMT system. IRSTLM (Federico et al., 2008) was used to train 5-gram language models over the target side of the (selected) bilingual corpora. …

System for Automatic Transcription of Sessions of the Polish Senate K MARASEK, D KORŽINEK, ? BROCKI – Archives of Acoustics, 2014 – acoustics.ippt.pan.pl … Such a model can be used in practice, even though it is still quite large, consuming as much as 1 GB of memory. Several programs for N-gram model creation were tested: SRILM (Stolcke et al., 2002), IRSTLM (Federico et al., 2008), and MITLM (Glass et al., 2009). …

Dynamic Models in Moses for Online Adaptation N Bertoldi – The Prague Bulletin of Mathematical Linguistics, 2014 – ufal.mff.cuni.cz … The target language model, which belongs to the second group, is usually provided by a third-party toolkit, like SRILM (Stolcke, 2002), IRSTLM (Fed- erico et al., 2008), randLM (Talbot and Osborne, 2007), kenLM (Heafield et al., 2013); however, a basic LM implementation is … Cited by 1 Related articles All 5 versions

An intelligent sample selection approach to language model adaptation for hand-written text recognition J Tanha, J de Does, K Depuydt – people.sabanciuniv.edu … Another approach is to allow the user to set the probability for each unknown word, as in the IRSTLM toolkit [7]. Neither of these solve the problem indicated above. … Irstlm: an open source toolkit for handling large scale language models. In Interspeech, pages 1618– 1621, 2008. …

Learning Hierarchical Translation Spans J Zhang, M Utiyama, E Sumita, H Zhao – aclweb.org … It is true that because of multiple parallel sen- tences, a source span can be applied with transla- 3http://hlt.fbk.eu/en/irstlm 186 Page 5. tion rules in one sentence pair but not in another sentence pair. So we used the probability score as a feature in the decoding. …

Learning Alternative Name Spellings J Sukharev, L Zhukov, A Popescul – arXiv preprint arXiv:1405.2048, 2014 – arxiv.org … things. 2. Language model training. A language model is a set of statistics gen- erated for an n-gram representation built with the target language. We used IRSTLM [13], a statistical language model tool for this pur- pose. As … Related articles All 2 versions

Cross Lingual Snippet Generation Using Snippet Translation System P Lohar, P Bhaskar, S Pal… – … Linguistics and Intelligent …, 2014 – Springer … For these reasons, translating snippets is more difficult than translating nor- mal sentences. Though the snippets may not be the container of well-formed sen- tences, most of the times their meaning can be perceived by the readers. 4 http://sourceforge.net/projects/irstlm/ Page 8. … Cited by 1 Related articles

Enhancing Statistical Machine Translation with Bilingual Terminology in a CAT Environment M Arcan, M Turchi, S Tonelli, P Buitelaar – Proceedings of the 11th …, 2014 – matecat.com … For each translation task, we use the statistical translation toolkit Moses (Koehn et al., 2007), where the word alignments were built with the GIZA++ toolkit (Och and Ney, 2003). The IRSTLM toolkit (Federico et al., 2008) was used to build the 5-gram language model. … Cited by 1

Task-based Evaluation of the PANACEA Production Chain V Aleksic, C Schwarz, G Thurmair, P Prokopidis… – 2014 – repositori.upf.edu … the Moses recaser. The lowercased versions of the target sides are used for training an interpolated 5- gram language model with Kneser-Ney discounting using the IRSTLM toolkit (Federico et al. 2011). Translation models … Cited by 1 Related articles All 8 versions

Translating without in-domain corpus: Machine translation post-editing with online learning techniques AL Lagarda, D Ortiz-Martínez, V Alabau… – Computer Speech & …, 2014 – Elsevier Globalization has dramatically increased the need of translating information from one language to another. Frequently, such translation needs should be satisfie.

An English to Xitsonga statistical machine translation system for the government domain CA McKellar – prasa.org … Czech Republic, 2007, pp. 177-180. [6] NBM Federico and M. Cettolo, “Irstlm: an open source toolkit for handling large scale language models,” in Proceedings of Interspeech, Brisbane, September 2008. [7] W. Pienaar and …

IIIT-H System Submission for FIRE2014 Shared Task on Transliterated Search IABVM Aniruddha, TRABM Shrivastava – isical.ac.in Page 1. IIIT-H System Submission for FIRE2014 Shared Task on Transliterated Search Irshad Ahmad Bhat Vandan Mujadia Aniruddha Tammewar Riyaz Ahmad Bhat Manish Shrivastava Language Technologies Research Centre …

Qualitative: Open source Python tool for Quality Estimation over multiple Machine Translation outputs A Eleftherios, P Lukas, S Sven – The Prague Bulletin of Mathematical …, 2014 – degruyter.com … In Principles of Data Mining and Knowledge Dis- covery, pages 537–539, 2004. Federico, Marcello, Nicola Bertoldi, and Mauro Cettolo. IRSTLM: an open source toolkit for handling large scale language models. In Interspeech, pages 1618–1621. ISCA, 2008. …

Selection of correction candidates for the normalization of Spanish user-generated content M MELERO, MR COSTA-JUSSÀ… – Natural Language … – Cambridge Univ Press Page 1. Natural Language Engineering: page 1 of 27. c Cambridge University Press 2014 doi:10.1017/S1351324914000011 1 Selection of correction candidates for the normalization of Spanish user-generated content M. MELERO1 … Related articles All 2 versions

Phrase table pruning by modeling the content of phrases F Azadi, S Khadivi – … (IST), 2014 7th International Symposium on, 2014 – ieeexplore.ieee.org … with some of the other popular pruning methods. These includes count-based and significance 1 https://github.com/moses-smt/mosesdecoder 2 http://sourceforge.net/ projects/irstlm/ pruning, that are trying to identify and prune …

Is getting the right answer just about choosing the right words? The role of syntactically-informed features in short answer scoring D Higgins, C Brew, M Heilman, R Ziai, L Chen… – arXiv preprint arXiv: …, 2014 – arxiv.org Page 1. Is getting the right answer just about choosing the right words? The role of syntactically-informed features in short answer scoring. Derrick Higgins*, Chris Brew†, Michael Heilman*, Ramon Ziai‡, Lei Chen*, Aoife Cahill … Cited by 1 Related articles All 2 versions

Quality estimation-guided supplementary data selection for domain adaptation of statistical machine translation P Banerjee, R Rubino, J Roturier, J van Genabith – Machine Translation, 2014 – Springer … 123 Page 12. P. Banerjee et al. development set (devset) optimised on BLEU (Papineni et al. 2002). The IRSTLM toolkit (Federico et al. 2008) is used for training all the 5-gram language models as well as for learning the linear interpolation weights using EM. …

Online adaptation to post-edits for phrase-based statistical machine translation N Bertoldi, P Simianer, M Cettolo, K Wäschle… – Machine …, 2014 – Springer … settings. The global 5-g LM is smoothed by the improved Kneser–Ney technique, and estimated on the target monolingual side of the parallel training data using the IRSTLM toolkit (Federico et al. 2008). Models are case- sensitive. … Cited by 3

Pipeline Creation Language for Machine Translation I Johnson – The Prague Bulletin of Mathematical Linguistics, 2014 – degruyter.com Page 1. The Prague Bulletin of Mathematical Linguistics NUMBER 101 APRIL 2014 55–69 Pipeline Creation Language for Machine Translation Ian Johnson Capita Translation and Interpreting Abstract A pipeline is a commonly used architecture in machine translation (MT). … Related articles All 4 versions

Weighted Finite-State Methods for Spell-Checking and Correction T Pirinen – 2014 – helda.helsinki.fi Page 1. Department of Modern Languages 2014 Weighted Finite-State Methods for Spell-Checking and Correction Tommi A Pirinen Academic dissertation to be publicly discussed, by due permission of the Fac- ulty of Arts at … Cited by 1 Related articles All 3 versions

Improving Statistical Machine Translation Through Adaptation And Learning JBM Acebal – 2014 – tdx.cat Page 1. Improving Statistical Machine Translation Through Adaptation And Learning Carlos A. Henríquez Q. Thesis advisor: Prof. Dr. José B. Mariño Acebal Thesis co-advisor: Rafael E. Banchs. Thesis presented to obtain the … Related articles

Improving statistical machine translation through adaptation and learning Q Henriquez, A Carlos – 2014 – tesisenxarxa.net Page 1. Improving Statistical Machine Translation Through Adaptation And Learning Carlos A. Henríquez Q. Thesis advisor: Prof. Dr. José B. Mariño Acebal Thesis co-advisor: Rafael E. Banchs. Thesis presented to obtain the …

Discourse in Statistical Machine Translation C Hardmeier – 2014 – diva-portal.org Page 1. ACTA UNIVERSITATIS UPSALIENSIS Studia Linguistica Upsaliensia 15 Page 2. Page 3. Discourse in Statistical Machine Translation Christian Hardmeier Page 4. Dissertation presented at Uppsala University to be publicly … Cited by 3 Related articles