Pivot Language for Low-Resource Machine Translation

Abhimanyu Talwar; Julien Laasri

REVIEW 5 major objections 6 minor 12 references

Pivot Language for Low-Resource Machine Translation

T0 review · 5 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Routing Nepali through a Hindi pivot lifts fully supervised translation from 7.6 to 14.2 SacreBLEU on the devtest set.

desk verdict A readable course project with a new Nepali-Hindi corpus and a plausible transfer result, but the headline 6.6-point gain over the prior baseline is not controlled and the artifacts are missing. read the letter →

arxiv 2505.14553 v2 pith:ZKUJEXNV submitted 2025-05-20 cs.CL cs.LG

Abhimanyu Talwar , Julien Laasri This is my paper

classification cs.CLcs.LG

keywords low-resourcemachinetranslationpivotlanguagetransfermethodbacktranslationNepaliHindiSacreBLEUTransformer

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

The paper sets out to show that a related, higher-resource language can substitute for missing parallel data in low-resource machine translation. Using Hindi as a pivot, the authors assemble a 284K-sentence Nepali-Hindi corpus, combine it with a 1.49M-sentence Hindi-English corpus, and train two Transformers that are chained at inference: Nepali to Hindi, then Hindi to English. This fully supervised transfer pipeline scores 14.2 SacreBLEU on the Nepali-English devtest set, 6.6 points above the fully supervised baseline. A semi-supervised variant that backtranslates Hindi monolingual data into the Nepali-Hindi pair reaches only 10.7, which the paper attributes to synthetic source noise eroding the natural similarity between Nepali and Hindi. The paper presents this as a practical blueprint for low-resource pairs that have a close high-resource relative.

What carries the argument

The carrying mechanism is the transfer method with $n=m=1$: at inference, the source sentence is decoded once by a Nepali-to-Hindi Transformer to produce one Hindi hypothesis, and that hypothesis is decoded once by a Hindi-to-English Transformer to produce the final English sentence, eliminating any rescoring stage. The method leans on two properties: the close structural relatedness of Nepali and Hindi (same Devanagari script, high lexical overlap, shared SOV word order and gendered morphology, so word reordering between them is minimal) and the availability of a large, diverse Hindi-English corpus of about 1.49M sentences plus a newly assembled 284K-sentence Nepali-Hindi corpus from three sources. All models are Transformers with the hyperparameters of the baseline and joint byte-pair-encoding vocabularies of 5,000 symbols.

What would settle it

Train the same fully supervised transfer system five times with different random seeds and compare the spread of devtest SacreBLEU scores against the 7.6 baseline; if any run lands at or below 7.6, or the gap is within the run-to-run variance, the claimed 6.6-point improvement is not established.

Watch

Extended reading notes

Core claim

The paper's central claim is that for Nepali-to-English translation, routing the translation through Hindi as a pivot—training separate Nepali-to-Hindi and Hindi-to-English Transformers and chaining them with a single candidate each—yields a devtest SacreBLEU score of 14.2, exceeding the fully supervised direct baseline of 7.6 reported by Guzmán et al. (2019) by 6.6 points. A semi-supervised variant that applies one iteration of backtranslation to the Nepali-Hindi pair scores 10.7, below the semi-supervised baseline of 15.1; the paper attributes this to backtranslation injecting noise into a source pair that was already easy because of Hindi-Nepali similarity. The paper also argues that Hindi is a suitable pivot because of its historical and lexical overlap with Nepali, shared Devanagari script, matching SOV word order, and shared morphological features such as gender agreement.

Load-bearing premise

The load-bearing premise is that the single reported 14.2 devtest SacreBLEU score is a reliable, comparable estimate—that the 6.6-point gap over the Guzmán et al. baseline reflects the Hindi-pivot method rather than the extra 1.77 million training sentences, different preprocessing, or run-to-run variation.

Editorial extensions

If this is right

A fully supervised pivot chain can beat direct fully supervised training for a low-resource pair without adding any new source-target parallel text.
End-to-end English quality is bounded by the pivot-to-target leg; the paper's Hindi-to-English model stopped at 55 epochs and the authors expect more training to raise the final score.
Backtranslation can hurt on a closely related pair: adding one iteration to Nepali-Hindi dropped the devtest score from 14.2 to 10.7, below the semi-supervised baseline of 15.1.
The transfer method is portable, since the paper frames it as applicable to other low-resource pairs and identifies the data requirements for building the two legs.

Reading between the lines

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

My inference: the 6.6-point gain conflates the pivot method with extra training data, because the transfer system draws on 1.49M Hindi-English and 284K Nepali-Hindi parallel pairs that the direct baseline never saw; a direct baseline trained on the same total data would isolate the pivot's contribution.
My inference: the reported gain is specific to the devtest split; the paper's own dev-set score for the fully supervised transfer model is 11.3 and no dev-set baseline is given, so the advantage may not be uniform across evaluation sets.
My inference: the same relatedness criteria—shared script, matching word order, and morphological overlap—could select pivots for other low-resource Indic pairs, such as Sinhala-English, making the recipe testable beyond Nepali.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Worth a look if you work on Nepali or Hindi MT, but don't take the headline at face value. The paper assembles a new Nepali-Hindi parallel corpus (OPUS, Bible, ILTPDC, 284K pairs) and applies the standard transfer method with n=m=1: a Nepali-Hindi Transformer and a Hindi-English Transformer, concatenated at inference. That is a legitimate extension, and the specific result—14.2 SacreBLEU on FLORES Ne→En devtest—is new. The authors also write clearly about why Hindi is a plausible pivot (shared script, SOV order, gendered morphology) and are honest about the semi-supervised backtranslation experiment hurting (10.7 vs 15.1 baseline).

The soft spot is the central comparison. The stress-test note's three points hold: the 7.6 baseline is copied from Guzmán et al. and was trained only on the 569K Nepali-English parallel corpus; the transfer pipeline consumes 1.49M Hindi-English pairs plus 284K Nepali-Hindi pairs, so the +6.6 could just be more parallel data and different domains, not pivoting. Table 5 says the authors' scores are detokenized SacreBLEU, but no one has shown Guzmán's 7.6 was computed with the same SacreBLEU version; a tokenization mismatch alone can move BLEU by several points. And the 14.2 is from one run—no seeds, no variance. The authors themselves invoke noise in Section 8 to explain the semi-supervised drop, which cuts against confidence in the main score. I also note that the Hindi-English model only ran 55 of 100 epochs (Section 6.2), so the reported transfer score is from an undertrained component.

No code, data, or training scripts are released, so the new corpus cannot be reused. The paper is essentially a readable, well-motivated course project (it says CS 287 Final Project) with an unverified headline result.

Who this is for: low-resource MT practitioners working with Hindi-Nepali–related pairs might get a useful pointer, but the lack of artifacts limits practical value. I would not cite it in the next year. As an editor, I would desk reject at a main venue; a workshop might give useful feedback if the authors supply artifacts and reproduce the baseline under the same pipeline. As it stands, the central claim is not established.

Referee Report

5 major / 6 minor

Summary. The paper proposes using Hindi as a pivot language for Nepali-to-English machine translation. The authors construct a Nepali-Hindi parallel corpus (~284K sentence pairs) from OPUS, Bible, and ILTPDC sources; use the IIT Bombay Hindi-English parallel corpus (~1.49M pairs); and apply two transfer-method variants: a fully-supervised pipeline that translates Nepali to Hindi and then Hindi to English with n=m=1, and a semi-supervised pipeline that inserts one iteration of backtranslation into the Nepali-Hindi leg. The main claim is that the fully-supervised transfer method achieves 14.2 detokenized SacreBLEU on the Nepali-English devtest set, which is 6.6 points higher than the fully-supervised baseline of 7.6 reported by Guzmán et al. (2019), while the semi-supervised variant reaches 10.7, below the 15.1 semi-supervised baseline. The paper also discusses linguistic reasons why Hindi is a suitable pivot, including shared script, word order, and morphology.

Significance. If the headline comparison were controlled, the result would be a useful contribution to low-resource MT: it demonstrates that a simple Hindi-pivot transfer, using relatively small additional parallel data, can outperform a direct supervised baseline on the Nepali-English pair, and it provides a new Nepali-Hindi corpus (~284K pairs) that the community could reuse. The paper is also transparent about the evaluation metric (SacreBLEU for English outputs, tokenized BLEU for Nepali-Hindi) and openly notes that the Hindi-English model was trained for only 55 of 100 epochs. It fails, however, to establish that the 6.6-point gain is due to pivoting rather than to the additional 1.49M Hindi-English and 284K Nepali-Hindi training pairs, different preprocessing, or run-to-run variation; the absence of a rerun of the Guzmán et al. baseline in the same pipeline, of variance estimates, and of code/data makes the central claim unverifiable as presented. These are experimental-rigor issues that can be addressed within the manuscript's scope, but they must be fixed before publication.

major comments (5)

[§7, Table 5] The 6.6-point improvement of 'OUR-Transfer Method (Fully-Supervised)' over the 'BASELINE (Fully-Supervised)' is not a controlled comparison. The baseline 7.6 is imported from Guzmán et al. (2019), while the transfer pipeline consumes 284K Nepali-Hindi pairs (Table 1) and 1.49M Hindi-English pairs (Table 2) that the baseline never saw, and the preprocessing is described only as 'similar' to Guzmán et al. (§6.1). To support the claim that pivoting through Hindi is responsible for the gain, the authors must either rerun the fully-supervised direct Nepali-English baseline under their own preprocessing and evaluation pipeline, or provide an ablation that trains the transfer pipeline on comparable amounts of direct data.
[§7, Table 5; §8] The reported devtest score of 14.2 is from a single training run, with no seeds, no standard deviation, and no significance test. The paper itself shows that such comparisons are noisy: §8 attributes the 2.4-point drop from 14.2 to 10.7 in the semi-supervised variant to harmful backtranslation and related noise, and run-to-run variance in low-resource NMT is substantial. The authors should report multiple runs (at least 3) and confidence intervals for both the transfer and baseline scores.
[§6.2; §9] The Hindi-to-English Transformer stopped at 55 of 100 epochs 'due to time constraints,' and §9 itself states that the score 'can be improved simply by letting our Hindi-English Transformer train for a few more days until we reach convergence.' The headline 14.2 is therefore an under-trained snapshot, not a converged or representative estimate. The paper should state whether the 47.2 and 16.8 intermediate scores are from checkpoints at 55 epochs and, if so, add a converged checkpoint or at least show a learning curve so the reader can judge how much the final score could change.
[§7, Table 5 footnote] The paper reports 'detokenized SacreBLEU' for {Ne,Hi}→En, but the Guzmán et al. baseline of 7.6 is simply taken from the published table, and no evidence is given that the two numbers were produced with the same SacreBLEU version, tokenization, and detokenization settings. Since SacreBLEU scores can differ by several points across versions and settings, the authors must recompute the baseline with the same evaluation script (and ideally the same preprocessing) before claiming a 6.6-point gap.
[§2.1; §6] The paper does not provide code or data for the corpus-construction pipeline (cleaning, deduplication, splitting of the Nepali-Hindi corpus) or for the training/evaluation pipeline, so the reported 14.2 cannot be independently reproduced. For a claim that rests on a single numerical comparison, this is a load-bearing reproducibility gap; at minimum, the corpus statistics and all preprocessing scripts should be released.

minor comments (6)

[Abstract; §1; §7] The abstract and Section 1 refer to a 'devtest Set SacreBLEU score of 14.2' and a 'Test Set BLEU score of 14.2,' while the footnote to Table 5 clarifies that the English-side scores are detokenized SacreBLEU and the Nepali-Hindi scores are tokenized BLEU; the terminology should be aligned throughout.
[§7, Table 5] Table 5 shows a 'Ne → En dev' column with no baseline entries; the paper should state why dev scores are not compared and how the dev/devtest split relates to Guzmán et al.'s official split.
[§4.2; §8] Section 4.2 says 300K Hindi sentences were randomly subsampled for backtranslation, but Section 8 does not report the quality of the backtranslated Nepali sentences; adding the BLEU of the backtranslation model and a few sample outputs would make the 'noise' explanation more convincing.
[§2.1, Table 1] The paper does not specify how the four Nepali-Hindi sources (GNOME/KDE/Ubuntu, Bible, Agriculture/Entertainment) were cleaned, deduplicated, and sentence-aligned; these details are needed to reuse the corpus.
[§3.3] The formula uses the intersection 'TEH ∩ THN' without defining the phrase-table notation; the reader has to infer that TEH and THN are sets of phrase pairs from Hindi-English and Nepali-Hindi, respectively.
[§5.2] The claim that 'a high amount of reordering is required' is supported by one anecdotal example; a quantitative reordering metric over the corpora would be more compelling.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the transfer result is evaluated on an external devtest set against a published baseline, with no fitted parameter renamed as a prediction and no load-bearing self-citation chain.

full rationale

The paper's derivation chain is not circular. The headline claim, a 14.2 SacreBLEU devtest score for the Hindi-pivot transfer method versus a 7.6 fully-supervised baseline from Guzmán et al. (2019), is an experimental measurement evaluated on the externally provided devtest set using a standard metric (SacreBLEU, Post 2018). Nothing in the paper defines the reported score in terms of the baseline, fits a parameter to the devtest and then re-predicts it, or invokes a result by the same authors to force the outcome. The pivot-language choice is justified by linguistic relatedness and data availability, not by a self-citation. The cited external works (Guzmán et al. 2019, Wu & Wang 2009, Sennrich et al. 2015, Kunchukuttan et al. 2017) supply baselines, methods, and data, but the central evaluation is independent of the authors' own prior claims. Concerns raised in the paper about single-run variance, the hindi-English model training for only 55 of 100 epochs, and the transfer system consuming additional parallel data are legitimate threats to the strength and comparability of the 6.6-point gain, but they concern experimental rigor and baseline comparability, not definitional or self-citational circularity. No circular step can be exhibited, so the appropriate score is 0.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

No parameters are fitted to the reported BLEU score. The main free choices are the BPE vocabulary size and the backtranslation subsample size; both are design decisions. The evaluation depends on four domain assumptions: that SacreBLEU is a sufficient measure, that the assembled corpora are clean and representative, that hyperparameters transfer across language pairs, and that Hindi-Nepali relatedness makes transfer beneficial. None of these is independently validated.

free parameters (2)

Joint BPE vocabulary size = 5000
Section 6.1; chosen following Guzmán et al. Affects all models, but is a standard preprocessing choice rather than fitted to the test score.
Random Hindi monolingual subsample size for backtranslation = 300000
Section 4.2; chosen for the semi-supervised variant. Affects the semi-supervised BLEU score, which is a secondary claim.

assumptions (4)

domain assumption SacreBLEU on the Guzmán et al. devtest set is a valid and sufficient measure of translation quality.
The paper uses only SacreBLEU in Section 7 and does not report human evaluation, confidence intervals, or significance tests.
domain assumption The assembled Nepali-Hindi corpus from OPUS, Bible, and ILTPDC is clean and diverse enough to train a reliable pivot model.
Section 2.1 describes the data combination but not the cleaning procedure; the pivot model quality depends on this corpus.
domain assumption Hyperparameters from Guzmán et al. transfer appropriately to Nepali-Hindi and Hindi-English pairs.
Section 6.2 reuses the baseline hyperparameters without sensitivity analysis or tuning for the new language pairs.
domain assumption Hindi and Nepali are related enough that a two-step transfer through Hindi helps rather than compounds errors.
Section 5 argues relatedness, but the empirical comparison to a direct model is not controlled, so this remains an assumption.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Pivot Language for Low-Resource Machine Translation." pith.science (2026). https://pith.science/paper/ZKUJEXNV

@misc{pith2026250514553,
  author       = {Pith},
  title        = {Pith review of: Pivot Language for Low-Resource Machine Translation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZKUJEXNV}},
  note         = {Machine review of arXiv:2505.14553}
}

read the original abstract

Certain pairs of languages suffer from lack of a parallel corpus which is large in size and diverse in domain. One of the ways this is overcome is via use of a pivot language. In this paper we use Hindi as a pivot language to translate Nepali into English. We describe what makes Hindi a good candidate for the pivot. We discuss ways in which a pivot language can be used, and use two such approaches - the Transfer Method (fully supervised) and Backtranslation (semi-supervised) - to translate Nepali into English. Using the former, we are able to achieve a devtest Set SacreBLEU score of 14.2, which improves the baseline fully supervised score reported by (Guzman et al., 2019) by 6.6 points. While we are slightly below the semi-supervised baseline score of 15.1, we discuss what may have caused this under-performance, and suggest scope for future work.

Figures

Figures reproduced from arXiv: 2505.14553 by the authors.

**Figure 1.** Examples of ill-formed sentences from the NepaliEnglish Parallel Corpus. Dataset Train Valid Test GNOME/KDE/Ubuntu 213K 464 2344 Bible 30K 65 330 Agriculture and Entertainment Domain 41K 87 444 Total 284K 616 3118 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗

**Figure 2.** The amount of word reordering required is minimal when translating between Hindi and Nepali but is significant when translating between Nepali and English. along these dimensions below. 5.1. Historical Relatedness Nepal shares part of its border with northern India and by some estimates, nearly 40% of Nepal’s population also speaks Hindi. Both languages are written in the same script of Devanagari, and if someone ca… view at source ↗

**Figure 3.** Nepali and Hindi share several morphological features and are different from English. is different when we are talking about the female gender compared to when we talk about the male gender. Another example in [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

12 extracted references · 4 canonical work pages

[6]

org/abs/1710.02855

URL http://arxiv. org/abs/1710.02855. Lample, Guillaume, Ott, Myle, Conneau, Alexis, Denoyer, Ludovic, and Ranzato, Marc’Aurelio. Phrase-based & neural unsupervised machine translation. CoRR, abs/1804.07755,

arXiv
[9]

org/abs/1804.08771

URL http://arxiv. org/abs/1804.08771. Sennrich, Rico, Haddow, Barry, and Birch, Alexandra. Im- proving neural machine translation models with mono- lingual data. CoRR, abs/1511.06709,

arXiv
[10]

Sinha, Prajadhip, Sarma, Bairab, and Shyam Purkayastha, Bipul

URLhttp: //arxiv.org/abs/1511.06709. Sinha, Prajadhip, Sarma, Bairab, and Shyam Purkayastha, Bipul. Kinship terms in nepali language and its morphol- ogy. International Journal of Computer Applications , 58, 10

arXiv
[12]

Tiedemann, J¨org

URL http://arxiv.org/ abs/1512.00567. Tiedemann, J¨org. Parallel data, tools and interfaces in opus. In Chair), Nicoletta Calzolari (Conference, Choukri, Khalid, Declerck, Thierry, Do ˘gan, Mehmet U ˘gur, Mae- gaard, Bente, Mariani, Joseph, Moreno, Asuncion, Odijk, Jan, and Piperidis, Stelios (eds.), Proceedings of the Eight International Conference on La...

arXiv
[13]

ISBN 978-2-9517408-7-7

European Language Resources Association (ELRA). ISBN 978-2-9517408-7-7. CS 287 Final Project Utiyama, Masao and Isahara, Hitoshi. A comparison of pivot methods for phrase-based statistical machine translation. In Human Language Technologies 2007: The Conference of the North American Chapter of the Association for Computational Linguistics; Proceed- ings o...

work page 2007
[15]

Wu, Hua and Wang, Haifeng

URL http://arxiv.org/ abs/1706.03762. Wu, Hua and Wang, Haifeng. Revisiting pivot language approach for machine translation. In Proceedings of the Joint Conference of the 47th Annual Meeting of the ACL and the 4th International Joint Conference on Natural Language Processing of the AFNLP: V olume 1 - V ol- ume 1 , ACL ’09, pp. 154–162, Stroudsburg, PA, USA,

arXiv
[17]

doi: 10.18653/v1/ D17-1319

doi:10.18653/v1/
[2008]

URL http://dl.acm.org/ citation.cfm?id=1613715.1613809

Association for Compu- tational Linguistics. URL http://dl.acm.org/ citation.cfm?id=1613715.1613809. Bojar, Ondrej, Diatka, V ojt ˇech, Rychl ´y, Pavel, Stranak, Pavel, Suchomel, Vit, Tamchyna, Ale ˇs, and Zeman, Daniel. HindEnCorp - Hindi-English and Hindi-only corpus for machine translation. In Proceedings of the Ninth International Conference on Langua...

work page arXiv

Show all 12 references

[2009]

ISBN 978-1-932432-45-9

Association for Computational Linguistics. ISBN 978-1-932432-45-9. URL http://dl.acm.org/ citation.cfm?id=1687878.1687902. Xu, Hainan and Koehn, Philipp. Zipporah: a fast and scal- able data cleaning system for noisy web-crawled paral- lel corpora. pp. 2945–2950, 01
[2017]

Christodouloupoulos, Christos and Steedman, Mark

URL http: //arxiv.org/abs/1709.08898. Christodouloupoulos, Christos and Steedman, Mark. A massively parallel corpus: the bible in 100 languages. Language Resources and Evaluation , 49(2):375–395, Jun

arXiv
[2018]

Papineni, Kishore, Roukos, Salim, Ward, Todd, and jing Zhu, Wei

URL http://arxiv.org/ abs/1804.07755. Papineni, Kishore, Roukos, Salim, Ward, Todd, and jing Zhu, Wei. Bleu: a method for automatic evaluation of machine translation. pp. 311–318,

arXiv
[2019]

Kunchukuttan, Anoop, Mehta, Pratik, and Bhattacharyya, Pushpak

URL http://arxiv.org/ abs/1902.01382. Kunchukuttan, Anoop, Mehta, Pratik, and Bhattacharyya, Pushpak. The IIT bombay english-hindi parallel corpus. CoRR, abs/1710.02855,

1902 arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [6]

org/abs/1710.02855

URL http://arxiv. org/abs/1710.02855. Lample, Guillaume, Ott, Myle, Conneau, Alexis, Denoyer, Ludovic, and Ranzato, Marc’Aurelio. Phrase-based & neural unsupervised machine translation. CoRR, abs/1804.07755,

arXiv

[2] [9]

org/abs/1804.08771

URL http://arxiv. org/abs/1804.08771. Sennrich, Rico, Haddow, Barry, and Birch, Alexandra. Im- proving neural machine translation models with mono- lingual data. CoRR, abs/1511.06709,

arXiv

[3] [10]

Sinha, Prajadhip, Sarma, Bairab, and Shyam Purkayastha, Bipul

URLhttp: //arxiv.org/abs/1511.06709. Sinha, Prajadhip, Sarma, Bairab, and Shyam Purkayastha, Bipul. Kinship terms in nepali language and its morphol- ogy. International Journal of Computer Applications , 58, 10

arXiv

[4] [12]

Tiedemann, J¨org

URL http://arxiv.org/ abs/1512.00567. Tiedemann, J¨org. Parallel data, tools and interfaces in opus. In Chair), Nicoletta Calzolari (Conference, Choukri, Khalid, Declerck, Thierry, Do ˘gan, Mehmet U ˘gur, Mae- gaard, Bente, Mariani, Joseph, Moreno, Asuncion, Odijk, Jan, and Piperidis, Stelios (eds.), Proceedings of the Eight International Conference on La...

arXiv

[5] [13]

ISBN 978-2-9517408-7-7

European Language Resources Association (ELRA). ISBN 978-2-9517408-7-7. CS 287 Final Project Utiyama, Masao and Isahara, Hitoshi. A comparison of pivot methods for phrase-based statistical machine translation. In Human Language Technologies 2007: The Conference of the North American Chapter of the Association for Computational Linguistics; Proceed- ings o...

work page 2007

[6] [15]

Wu, Hua and Wang, Haifeng

URL http://arxiv.org/ abs/1706.03762. Wu, Hua and Wang, Haifeng. Revisiting pivot language approach for machine translation. In Proceedings of the Joint Conference of the 47th Annual Meeting of the ACL and the 4th International Joint Conference on Natural Language Processing of the AFNLP: V olume 1 - V ol- ume 1 , ACL ’09, pp. 154–162, Stroudsburg, PA, USA,

arXiv

[7] [17]

doi: 10.18653/v1/ D17-1319

doi:10.18653/v1/

[8] [2008]

URL http://dl.acm.org/ citation.cfm?id=1613715.1613809

Association for Compu- tational Linguistics. URL http://dl.acm.org/ citation.cfm?id=1613715.1613809. Bojar, Ondrej, Diatka, V ojt ˇech, Rychl ´y, Pavel, Stranak, Pavel, Suchomel, Vit, Tamchyna, Ale ˇs, and Zeman, Daniel. HindEnCorp - Hindi-English and Hindi-only corpus for machine translation. In Proceedings of the Ninth International Conference on Langua...

work page arXiv