Pith. sign in

REVIEW 3 major objections 4 minor 40 references

Transductive Data-Selection Algorithms for Fine-Tuning Neural Machine Translation

T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Pith's one-line read Fine-tuning a translation model on a small subset of sentences chosen for their n-gram overlap with the test set can beat both the generic base model and a model fine-tuned on the full in-domain data.

desk verdict Useful empirical comparison of transductive data selection for NMT fine-tuning, but the TF-IDF equations as written select the least similar sentences, so the paper's best result is not reproducible as specified. read the letter →

arxiv 1908.09532 v3 pith:QKWEIKD3 submitted 2019-08-26 cs.CL

classification cs.CL
keywords transductivedataselectionfine-tuningneuralmachinetranslationdomainadaptationfeaturedecayalgorithminfrequentn-gramrecoveryTF-IDFtest-time
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Translation quality often improves when a model is adapted to the document it will translate. This paper tests a transductive approach: use the test document itself to rank a pool of parallel sentences by source-side n-gram overlap, fine-tune a generic neural machine translation model on the top-ranked subset for one epoch, and translate. Across German-to-English news and health experiments, the selected subsets frequently outperform the base model and often outperform fine-tuning on the complete in-domain corpus. In the strongest result, a health-domain model fine-tuned on 500K sentences selected by TF-IDF from a mixture of general and in-domain data reaches BLEU 36.79, versus 34.69 for full-EMEA fine-tuning and 33.29 for the base model. The paper concludes that test-time adaptation with a small selected subset is a practical alternative to building separate domain models.

What carries the argument

The central mechanism is the transductive scoring function that maps each candidate sentence to a relevance score based on n-gram overlap with the test set. Three variants are used: TF-IDF cosine similarity (the maximum similarity to any test sentence), infrequent n-gram recovery (counting rare n-grams shared with the test set, with counts decayed as sentences are selected), and the feature decay algorithm, whose score for a sentence is the length-normalized sum $\sum_{ngr\in F_s} 0.5^{C_L(ngr)}$, where $C_L(ngr)$ is the number of times that n-gram already appears in the selected pool. The decay term encourages diversity, so the selected subset covers many different test-set n-grams rather than repeating the most similar sentences. The selected subset is then used to fine-tune the base model for one additional epoch.

What would settle it

Fine-tune the same base model on a random subset of the same size drawn from the same candidate pool, and compare BLEU on the test set with the transductive subsets; if the random subsets match the selected subsets' scores, the selection criteria add no value beyond reducing data size.

Watch

Extended reading notes

Core claim

The paper's central claim is that transductive data selection—ranking candidate parallel sentences by their source-side n-gram similarity to the test set—plus a single epoch of fine-tuning produces a model that translates that test set better than the generic model or a model fine-tuned on the full in-domain data. The claim is supported by experiments using three selection algorithms: TF-IDF cosine distance, infrequent n-gram recovery (INR), and the feature decay algorithm (FDA). All three select 100K–500K sentences from either the general training data, the in-domain corpus, or a mixture of both. In the news domain, the selected subsets recover the performance lost when fine-tuning on the full in-domain news corpus, and in the health domain the mixture experiment with TF-IDF achieves a statistically significant gain of 2.1 BLEU over full-EMEA fine-tuning (36.79 vs 34.69). The authors interpret this as evidence that the selected sentences are closer to the test document than the full domain corpus is.

Load-bearing premise

The load-bearing premise is that source-side n-gram overlap with the test set is a sufficient signal for choosing fine-tuning data; because no random-subset control is run, the gains attributed to selection could in principle come from the smaller fine-tuning data size alone.

Editorial extensions

If this is right

  • A translation provider that has access to the test document and owns the model can adapt at translation time without training or storing a separate model per domain.
  • Because only one epoch of fine-tuning is needed, the added latency is that of scoring the candidate pool and one training pass, making the approach feasible for document-level translation workflows.
  • The improvements are largest when the candidate pool mixes general and in-domain data, suggesting that adding a broad pool of sentences gives the selection algorithm more relevant material to find.
  • In the news experiments, selecting from the in-domain corpus alone still underperforms the base model, so the choice of candidate pool matters as much as the selection algorithm.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper never compares against a random subset of the same size, so part of the observed gain could come from fine-tuning on fewer, more predictable sentences rather than from the n-gram selection itself; a random-subset control would separate these effects.
  • The selection signals are purely source-side and lexical; incorporating target-side or semantic similarity (e.g., embeddings of test sentences) might further improve ranking, especially for test sets with little n-gram overlap.
  • Because all experiments use a single language pair and two domains, the generality of the conclusion across languages, domains, and model architectures remains an open question; the same protocol could be applied to multilingual models or low-resource settings.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes three transductive data-selection algorithms — TF-IDF distance, Infrequent n-gram Recovery (INR), and Feature Decay Algorithms (FDA) — that use the source side of a given test set to retrieve a subset of parallel sentences for fine-tuning a pre-trained NMT model. The authors evaluate German-to-English models on news and health test sets, comparing fine-tuning on selected subsets (100K/200K/500K sentences from general-domain, in-domain, or mixed candidate pools) against the base model and against full-data fine-tuning. The main claimed results are that fine-tuning on transductively selected subsets can outperform both the generic model and a domain-adapted model, with the strongest result being a TF-IDF-selected 500K mixture of BASE and EMEA data reaching 36.79 BLEU on the health test, versus 34.69 for full-EMEA fine-tuning (Table 11). Sections 7.1–7.3 also analyze the composition of selected data and the runtime/threshold behavior of INR.

Significance. If the results hold, the paper offers a practical recipe: given a test document at translation time, one can fine-tune an existing NMT model on a small, selected subset of available parallel data and outperform both a generic model and a model fine-tuned on the entire in-domain corpus, without training a full domain model. The work is also useful in framing data selection as a test-time adaptation tool and in comparing three selection criteria under a common experimental protocol. The paper ships a clean setup (WMT15 training data, OpenNMT-py, BPE, significance testing with multeval) and includes a useful analysis of the domain composition of the selected subsets (Table 9). However, the central empirical claim is currently weakened by a missing random-subset control and by an internal inconsistency in the definition of the TF-IDF method that produced the headline number; both issues must be resolved before the contribution is fully supported.

major comments (3)
  1. [Section 4, Eqs. (3)–(4)] The TF-IDF method as written does not implement the described transductive selection. Eq. (3) defines sim(a,b) = 1 - cos(w_a,w_b), which is cosine distance: identical vectors receive a score of 0 and orthogonal vectors receive a score of 1. Eq. (4) then scores each candidate by the maximum of this quantity over test sentences, i.e., it selects the candidate sentence with the largest distance to the test set, not the highest similarity. The surrounding text and the method name state the opposite. Since the strongest reported result (Table 11, TFIDF 500K, BLEU 36.79) is attributed to this method, the paper as written does not specify the algorithm that generated its central empirical claim. Please either correct Eq. (3) to use cosine similarity (remove the 1-) or change Eq. (4) to take the minimum, and explicitly state which variant was run in Section 7.3.
  2. [Sections 7.1–7.3, Tables 5–11] All comparisons of selected subsets are made against the full-data fine-tuned model or the base model, never against a random subset of the same size. This is a load-bearing control: without it, the observed improvements cannot be attributed to the selection algorithms, because fine-tuning on a smaller, possibly more homogeneous data set could by itself produce gains. The premise enters in every main experiment table (Tables 5–8, 10–11), where each selected subset at 100K/200K/500K should be accompanied by a random-subset baseline at the same size and from the same candidate pool. Please add these controls or state why they are unnecessary.
  3. [Tables 10 and 11] The first baseline column is labeled "BASE13" but contains the BASE12 values from Tables 1 and 2 (BLEU 26.16 on the news test and 33.29 on the health test, respectively). The text in Section 7.3 says this column "shows the model built performing 13 epochs," which is inconsistent with the values. This labeling error makes the claim that scores are "better than both baselines" ambiguous and should be corrected before publication.
minor comments (4)
  1. [Section 4, Eq. (5)] The text reads "the component max(0, t - CS(ngr)) is 0" but CS is not defined; it should presumably be CL(ngr). Also, clarify whether the infrequency threshold t is measured with respect to the selected pool L or the original candidate pool, since the current description makes Eq. (5) depend on an initially empty L.
  2. [Section 7.1, Tables 5 and 6] The text states that INR retrieves at most 230K sentences for news and 275K for health, but Tables 5 and 6 show INR rows for 200K and no rows for 500K; please state explicitly that the 500K INR experiments are absent because the retrieved count is below 500K.
  3. [Section 4, final paragraph] The description of the TF-IDF method calls Eq. (3) the "inverse of the cosine distance," which is not an accurate name for 1 - cos(wa,wb); if this wording follows the original work of Hildebrand et al. [14], please clarify the provenance and the intended sign convention.
  4. [Section 5, model configuration] The paper reports BPE with 89,500 merge operations and then says the vocabulary size is 50,000 words for each language; this could confuse readers about the effective subword vocabulary and should be clarified.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the transductive selection uses only source-side test text, evaluation uses independent references, and no fitted parameter is renamed as a prediction.

full rationale

The paper's chain of reasoning is empirical and self-contained. Data are selected by source-side n-gram signals (TF-IDF distance, infrequent n-gram recovery, feature decay) computed from the test source text, while quality is measured against human references that are never used in selection. The INR threshold t is chosen by runtime and retrieval-count criteria, not by tuning on test-set BLEU, and the subset sizes (100K/200K/500K) are fixed ahead of evaluation. The algorithms themselves are imported from external prior work ([14], [24], [4]), not from the authors' own papers; the self-citations that appear ([25]-[31]) concern related feature-decay and adaptation extensions and do not supply a load-bearing premise that forces the reported improvements. The central claim, that fine-tuning on a transductively selected subset can beat full-domain fine-tuning, is a measured result rather than a definitional equivalence. The skeptical observation that Eqs. (3)-(4) define a distance maximization (since sim = 1 - cos and the max is taken) is a possible correctness or implementation ambiguity, but it is not circularity: either variant is still evaluated against independent references, so the comparison does not reduce to the method's own inputs. Accordingly, no circular step is identified.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new theoretical entities and fits no parameters to test-set performance. The listed free parameters are configuration defaults or threshold choices based on computational constraints. The most load-bearing assumption is that source-side n-gram similarity identifies useful fine-tuning data, which is not tested against a random-subset baseline.

free parameters (7)
  • FDA decay parameter d = 0.5
    Default value from the original feature decay paper, used in Equation (6); results depend on it but it is not tuned to the test sets.
  • FDA exponent c = 0
    Default value in Equation (6), making the decay function depend only on the count in the selected pool.
  • FDA n-gram order = 3
    Chosen as the feature unit in Equations (6) and (7); no higher-order analysis is reported.
  • INR threshold t = 80 for news, 640 for health
    Chosen as the highest value satisfying a runtime limit of 48 hours or at least 500K retrieved sentences; this affects how many and which sentences are selected.
  • INR n-gram order = 3
    Used in Equation (5) to define the n-grams counted as infrequent.
  • Selected subset size = 100K, 200K, 500K lines
    Experimental factor, not fitted, but the results are specific to these subset sizes and to one-epoch fine-tuning.
  • Fine-tuning epoch count = 1 extra epoch
    The base model is fine-tuned for one additional epoch; more epochs might change the ranking of methods.
assumptions (5)
  • domain assumption Source-side n-gram overlap with the test set is a valid proxy for sentence relevance for fine-tuning.
    All three selection methods score candidate sentences by n-gram or vector similarity to the test source; the paper provides no independent evidence that this proxy identifies beneficial training sentences, and the missing random baseline leaves this assumption untested.
  • domain assumption One-epoch fine-tuning on a selected subset is a fair and sufficient adaptation procedure.
    The experiments fine-tune only for the 13th epoch; this is a modeling choice that could favor small subsets and is not varied across experiments.
  • domain assumption The BPE segmentation and OpenNMT-py default architecture are adequate for the comparison.
    The paper relies on 89500 BPE merge operations and a 2-layer LSTM with 500 hidden units from previous work; results may not transfer to other architectures or preprocessing choices.
  • standard math Bootstrap resampling significance tests at p=0.01 are reliable for the reported comparisons.
    The paper uses multeval with bootstrap resampling, a standard method in MT evaluation, but no multiple-testing correction is applied across the large number of table cells.
  • domain assumption The WMT15 general corpus, rapid2016 news corpus, EMEA health corpus, and Cochrane test set represent their intended domains.
    Domain representativeness is assumed; the paper notes that Europarl in the general data makes it closer to news, which may explain some domain-specific results.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Transductive Data-Selection Algorithms for Fine-Tuning Neural Machine Translation." pith.science (2026). https://pith.science/paper/QKWEIKD3

@misc{pith2026190809532,
  author       = {Pith},
  title        = {Pith review of: Transductive Data-Selection Algorithms for Fine-Tuning Neural Machine Translation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QKWEIKD3}},
  note         = {Machine review of arXiv:1908.09532}
}
read the original abstract

Machine Translation models are trained to translate a variety of documents from one language into another. However, models specifically trained for a particular characteristics of the documents tend to perform better. Fine-tuning is a technique for adapting an NMT model to some domain. In this work, we want to use this technique to adapt the model to a given test set. In particular, we are using transductive data selection algorithms which take advantage the information of the test set to retrieve sentences from a larger parallel set. In cases where the model is available at translation time (when the test set is provided), it can be adapted with a small subset of data, thereby achieving better performance than a generic model or a domain-adapted model.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 37 canonical work pages

  1. [1]

    Domain adaptation via pseudo in-domain data selection

    Amittai Axelrod, Xiaodong He, and Jian- feng Gao. Domain adaptation via pseudo in-domain data selection. In Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, pages 355– 362, Edinburgh, Scotland, UK., 2011

  2. [2]

    Meteor: An automatic metric for MT evaluation with improved correlation with human judgments

    Satanjeev Banerjee and Alon Lavie. Meteor: An automatic metric for MT evaluation with improved correlation with human judgments. In Proceedings of the ACL workshop on in- trinsic and extrinsic evaluation measures for machine translation and/or summarization , pages 65–72, Ann Arbor, Michigan, 2005

  3. [3]

    Feature decay algorithms for fast deployment of accurate statistical ma- chine translation systems

    Ergun Bic ¸ici. Feature decay algorithms for fast deployment of accurate statistical ma- chine translation systems. In Proceedings of the Eighth Workshop on Statistical Ma- chine Translation, pages 78–84, Sofia, Bul- garia, August 2013

  4. [4]

    Instance se- lection for machine translation using feature decay algorithms

    Ergun Bic ¸ici and Deniz Yuret. Instance se- lection for machine translation using feature decay algorithms. In Proceedings of the Sixth Workshop on Statistical Machine Trans- lation, pages 272–283, Edinburgh, Scotland, 2011

  5. [5]

    Findings of the 2017 conference on ma- chine translation (wmt17)

    Ond ˇrej Bojar, Rajen Chatterjee, Christian Federmann, Yvette Graham, Barry Had- dow, Shujian Huang, Matthias Huck, Philipp Koehn, Qun Liu, Varvara Logacheva, et al. Findings of the 2017 conference on ma- chine translation (wmt17). In Proceedings of the Second Conference on Machine Transla- tion, pages 169–214, Copenhagen, Denmark, 2017

  6. [6]

    Findings of the 2015 Work- shop on Statistical Machine Translation

    Ond ˇrej Bojar, Rajen Chatterjee, Christian Federmann, Barry Haddow, Matthias Huck, Chris Hokamp, Philipp Koehn, Varvara Lo- gacheva, Christof Monz, Matteo Negri, Matt Post, Carolina Scarton, Lucia Specia, and Marco Turchi. Findings of the 2015 Work- shop on Statistical Machine Translation. In Proceedings of the Tenth Workshop on Statis- tical Machine Tra...

  7. [7]

    Adapting neural ma- chine translation with parallel synthetic data

    Mara Chinea-Rios, Alvaro Peris, and Fran- cisco Casacuberta. Adapting neural ma- chine translation with parallel synthetic data. In Proceedings of the Second Conference on Machine Translation , pages 138–147, Copenhagen, Denmark, 2017

  8. [8]

    An empirical comparison of do- main adaptation methods for neural machine translation

    Chenhui Chu, Raj Dabre, and Sadao Kuro- hashi. An empirical comparison of do- main adaptation methods for neural machine translation. In Proceedings of the 55th An- nual Meeting of the Association for Com- putational Linguistics (V olume 2: Short Pa- pers), volume 2, pages 385–391, Vancouver, Canada, 2017

Show all 40 references
  1. [9]

    A survey of do- main adaptation for neural machine transla- tion

    Chenhui Chu and Rui Wang. A survey of do- main adaptation for neural machine transla- tion. arXiv preprint arXiv:1806.00258, 2018

  2. [10]

    Clark, Chris Dyer, Alon Lavie, and Noah A

    Jonathan H. Clark, Chris Dyer, Alon Lavie, and Noah A. Smith. Better hypothesis test- ing for statistical machine translation: Con- trolling for optimizer instability. In Pro- ceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language ...

  3. [11]

    Survey of data-selection methods in statistical ma- chine translation

    Sauleh Eetemadi, William Lewis, Kristina Toutanova, and Hayder Radha. Survey of data-selection methods in statistical ma- chine translation. Machine Translation, 29(3- 4):189–223, 2015

  4. [12]

    Fast domain adaptation for neural machine trans- lation

    Markus Freitag and Yaser Al-Onaizan. Fast domain adaptation for neural machine trans- lation. arXiv preprint arXiv:1612.06897 , 2016

  5. [13]

    Guillem Gasc ´o, Martha-Alicia Rocha, Germ´an Sanchis-Trilles, Jes ´us Andr ´es- Ferrer, and Francisco Casacuberta. Does more data always yield better translations? In Proceedings of the 13th Conference of the European Chapter of the Association for Computational Linguistics ,...

  6. [14]

    Adapta- tion of the translation model for statistical machine translation based on information re- trieval

    Almut Silja Hildebrand, Matthias Eck, Stephan V ogel, and Alex Waibel. Adapta- tion of the translation model for statistical machine translation based on information re- trieval. In Proceedings of the 10th Annual Conference of the European Association for Machine Translation ,...

  7. [15]

    Guillaume Klein, Yoon Kim, Yuntian Deng, Jean Senellart, and Alexander M. Rush. Opennmt: Open-source toolkit for neural ma- chine translation. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics-System Demon- strations, pages 67–72, Vanco...

  8. [16]

    Statistical significance tests for machine translation evaluation

    Philipp Koehn. Statistical significance tests for machine translation evaluation. In Pro- ceedings of the 2004 Conference on Empir- ical Methods in Natural Language Process- ing, pages 388–395, Barcelona, Spain, 2004

  9. [17]

    Europarl: A parallel cor- pus for statistical machine translation

    Philipp Koehn. Europarl: A parallel cor- pus for statistical machine translation. Ma- chine Translation Summit, 2005 , pages 79– 86, 2005

  10. [18]

    Binary codes capable of correcting deletions, insertions and rever- sals

    Vladimir Levenshtein. Binary codes capable of correcting deletions, insertions and rever- sals. In Soviet Physics Doklady , pages 707– 710, 1966

  11. [19]

    One sentence one model for neu- ral machine translation

    Xiaoqing Li, Jiajun Zhang, and Chengqing Zong. One sentence one model for neu- ral machine translation. In Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018), pages 910–917, Miyazaki, Japan, 2018

  12. [20]

    Lo- cally training the log-linear model for smt

    Lemao Liu, Hailong Cao, Taro Watanabe, Tiejun Zhao, Mo Yu, and Conghui Zhu. Lo- cally training the log-linear model for smt. In Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Lan- guage Learning, pages 402...

  13. [21]

    Stanford neural machine translation systems for spoken language domains

    Minh-Thang Luong and Christopher D Man- ning. Stanford neural machine translation systems for spoken language domains. In Proceedings of the International Workshop on Spoken Language Translation , pages 76– 79, Da Nang, Vietnam, 2015

  14. [22]

    Distributed representations of words and phrases and their compositionality

    Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg S Corrado, and Jeff Dean. Distributed representations of words and phrases and their compositionality. In Advances in neural information processing systems, pages 3111– 3119, 2013

  15. [23]

    Bleu: a method for auto- matic evaluation of machine translation

    Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. Bleu: a method for auto- matic evaluation of machine translation. In Proceedings of 40th Annual Meeting of the Association for Computational Linguistics , pages 311–318, Philadelphia, Pennsylvania, USA, July 2002

  16. [24]

    Data selection for nmt using infrequent n-gram recovery

    Zuzanna Parcheta, Germ ´an Sanchis-Trilles, and Francisco Casacuberta. Data selection for nmt using infrequent n-gram recovery. In Proceedings of the 21st Annual Conference of the European Association for Machine Trans- lation, page 219–227, Alacant, Spain, 2018

  17. [25]

    Applying n-gram alignment entropy to improve feature decay algorithms

    Alberto Poncelas, Gideon Maillette de Buy Wenniger, and Andy Way. Applying n-gram alignment entropy to improve feature decay algorithms. The Prague Bulletin of Mathematical Linguistics , 108(1):245–256, 2017

  18. [26]

    Data selection with feature decay algorithms using an approximated target side

    Alberto Poncelas, Gideon Maillette de Buy Wenniger, and Andy Way. Data selection with feature decay algorithms using an approximated target side. In 15th International Workshop on Spoken Language Translation (IWSLT 2018) , pages 173–180, Bruges, Belgium, 2018

  19. [27]

    Fea- ture decay algorithms for neural machine translation

    Alberto Poncelas, Gideon Maillette de Buy Wenniger, and Andy Way. Fea- ture decay algorithms for neural machine translation. In Proceedings of the 21st An- nual Conference of the European Association for Machine Translation , pages 239–248, Alacant, Spain, 2018

  20. [28]

    Adap- tation of machine translation models with back-translated data using transductive data selection methods

    Alberto Poncelas, Gideon Maillette de Buy Wenniger, and Andy Way. Adap- tation of machine translation models with back-translated data using transductive data selection methods. In 20th International Conference on Computational Linguistics and Intelligent Text Processing, La R...

  21. [29]

    Adapting NMT to caption translation in Wikimedia Commons for low- resource languages

    Alberto Poncelas, Kepa Sarasola, Meghan Dowling, Andy Way, Gorka Labaka, and I˜naki Alegria. Adapting NMT to caption translation in Wikimedia Commons for low- resource languages. In 35th International Conference of the Spanish Society for Natu- ral Language Processing (SEPLN 2...

  22. [30]

    The ADAPT System Description for the IWSLT 2018 Basque to English Translation Task

    Alberto Poncelas, Andy Way, and Kepa Sara- sola. The ADAPT System Description for the IWSLT 2018 Basque to English Translation Task. In International Workshop on Spoken Language Translation, pages 72–82, Bruges, Belgium, 2018

  23. [31]

    Extending feature decay algorithms using alignment entropy

    Alberto Poncelas, Andy Way, and Antonio Toral. Extending feature decay algorithms using alignment entropy. In International Workshop on Future and Emerging Trends in Language Technology , pages 170–182, Seville, Spain, 2016

  24. [32]

    On the specification of term values in auto- matic indexing

    Gerard Salton and Chung-Shu Yang. On the specification of term values in auto- matic indexing. Journal of documentation , 29(4):351–372, 1973

  25. [33]

    Neural machine translation of rare words with subword units

    Rico Sennrich, Barry Haddow, and Alexan- dra Birch. Neural machine translation of rare words with subword units. In Proceedings of the 54th Annual Meeting of the Associa- tion for Computational Linguistics (V olume 1: Long Papers) , volume 1, pages 1715– 1725, Berlin, Germany, 2016

  26. [34]

    Extracting in- domain training corpora for neural machine translation using data selection methods

    Catarina Cruz Silva, Chao-Hong Liu, Alberto Poncelas, and Andy Way. Extracting in- domain training corpora for neural machine translation using data selection methods. In Proceedings of the Third Conference on Ma- chine Translation: Research Papers , pages 224–231, Brussels, B...

  27. [35]

    A study of translation edit rate with targeted human annotation

    Matthew Snover, Bonnie Dorr, Richard Schwartz, Linnea Micciulla, and John Makhoul. A study of translation edit rate with targeted human annotation. In Proceedings of the 7th Conference of the Association for Machine Translation in the Americas , pages 223–231, Cambridge, Massa...

  28. [36]

    News from OPUS - A collection of multilingual parallel corpora with tools and interfaces

    J ¨org Tiedemann. News from OPUS - A collection of multilingual parallel corpora with tools and interfaces. In N. Nicolov, K. Bontcheva, G. Angelova, and R. Mitkov, editors, Recent Advances in Natural Lan- guage Processing, volume V , pages 237–248. John Benjamins, Amsterdam/P...

  29. [37]

    Two methods for stabiliz- ing mert: Nict at iwslt 2009

    Masao Utiyama, Hirofumi Yamamoto, and Eiichiro Sumita. Two methods for stabiliz- ing mert: Nict at iwslt 2009. In International Workshop on Spoken Language Translation (IWSLT 2009) , pages 79–82, Tokyo, Japan, 2009

  30. [38]

    Dynamic data selection for neural machine translation

    Marlies van der Wees, Arianna Bisazza, and Christof Monz. Dynamic data selection for neural machine translation. In Proceedings of the 2017 Conference on Empirical Meth- ods in Natural Language Processing , pages 1400–1410, Copenhagen, Denmark, 2017

  31. [39]

    Vladimir N. Vapnik. Statistical Learning Theory. Wiley-Interscience, 1998

  32. [40]

    Findings of the wmt 2017 biomedical translation shared task

    Antonio Jimeno Yepes, Aur´elie N´ev´eol, Mar- iana Neves, Karin Verspoor, Ondrej Bojar, Arthur Boyer, Cristian Grozea, Barry Had- dow, Madeleine Kittner, Yvonne Lichtblau, et al. Findings of the wmt 2017 biomedical translation shared task. In Proceedings of the Second Conferen...

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.