Pith. sign in

REVIEW 4 major objections 5 minor 30 references

Lost in Transliteration: Bridging the Script Gap in Neural IR

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Fine-tuning multilingual rankers on an even mix of native and romanized queries nearly closes the retrieval gap between native and transliterated queries.

desk verdict Honest short paper with a genuinely new result for Russian, but the abstract's 'nearly same performance' doesn't hold for Chinese, and the whole recipe is only tested on canonical Uroman romanizations, not real user transliterations. read the letter →

arxiv 2505.08411 v1 pith:MTEU3CAJ submitted 2025-05-13 cs.IR

Andreas Chari , Iadh Ounis , Sean MacAvaney This is my paper · ORCID
classification cs.IR
keywords TransliterationNeuralinformationretrievalScriptgapMultilingualCross-scriptDenseRomanizationFine-tuning
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

Most search users in non-Latin-script languages sometimes type queries in Latin letters (Greeklish, Arabizi, pinyin-style Chinese), and this paper shows that current multilingual neural retrievers fail on those queries: BGE-M3's retrieval effectiveness drops by 97% for Chinese and 49% for Russian when queries are transliterated. The paper argues that this 'script gap' can be largely closed by fine-tuning on an even mixture of native and Uroman-romanized queries, while documents stay in the native script. With this 'transliterate-train' recipe, Russian transliterated queries reach nearly native-level performance, Chinese transliterated queries improve substantially, and native-script performance is preserved. The gains transfer to out-of-domain Russian collections but only partially to Chinese, where transliteration also flattens query nuance. If right, this is a cheap data-augmentation fix for a real usability problem.

What carries the argument

The central mechanism is the 'transliterate-train' procedure, an adaptation of translate-train: take a language's native training queries, romanize them with the Uroman tool, and fine-tune the ranker on a 50/50 mix of native and romanized queries paired with unchanged native-script documents. This exposes the model to pairs of surface forms for the same information need so that the embedding and reranking layers learn to treat the two scripts as referring to the same text. The argument largely rests on the measured performance gap between native and transliterated queries before and after this fine-tuning on mMARCO Chinese and Russian, plus out-of-domain checks on TREC NeuCLIR.

What would settle it

Take a set of queries actually typed in Latin letters by Russian and Chinese speakers (for example, from chat logs or real search sessions), transliterate the same queries with Uroman, and compare retrieval quality of the mixed-configuration models on both sets. If human-typed transliterations perform no better than the unfine-tuned baseline while the Uroman versions reproduce the paper's gains, the bridge does not extend to real user input.

Watch

Extended reading notes

Core claim

On the paper's own terms: fine-tuning a multilingual dense retriever (BGE-M3) and a multilingual cross-encoder reranker (mT5) on an even mix of native-script and transliterated queries, against native-script documents, lets the models match transliterated queries to relevant documents at nearly the same effectiveness as native queries in-domain. The mixed configuration is the only fine-tuning scheme tested that preserves native-text performance while significantly improving transliterated-query performance; training only on transliterations trades away native Chinese performance, and training only on native script does nothing for transliterations. Out-of-domain, the mixed model closes most of the Russian script gap but fails to do so for Chinese, and qualitative inspection shows romanized queries drift to generic results because transliteration flattens lexical nuance.

Load-bearing premise

The load-bearing premise is that Uroman's canonical romanizations adequately represent how real users type transliterated queries; if real Greeklish, Arabizi, or Latin-typed Chinese differ in spelling and style, the fine-tuning gains may not transfer to actual users.

Editorial extensions

If this is right

  • Romanized Russian queries can reach near-native retrieval quality after fine-tuning on mixed data, so users would not need to switch keyboards to get good results.
  • Existing multilingual dense retrievers and rerankers can be adapted with data augmentation alone; no new architecture is required.
  • The same augmentation is less effective when the native script and Latin share little surface similarity, so languages like Chinese need extra measures.
  • Out-of-domain collections may not see the full benefit, especially for Chinese; deployment should verify against target-domain queries.
  • Transliteration can blur query nuance (e.g., 'peanut allergy treatment' drifting to generic medicine), so script matching alone does not guarantee meaning preservation.

Reading between the lines

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

  • A direct test of the practical value would use human-typed transliterations from real interactions; Uroman's output is a canonical romanization, so real Greeklish, Arabizi, or pinyin-style queries with vowel drops and informal spellings may not reproduce the reported gains.
  • The authors' own qualitative results suggest the remaining Chinese out-of-domain gap is partly semantic ambiguity, so coupling transliterate-train with a script-alignment or contrastive objective could push generalization further.
  • Tokenizer behavior may be a lever: the authors report inconsistent results when Chinese queries were pre-tokenized before romanization, and the interaction between segmentation and romanization is not yet pinned down.
  • A production system could apply the same idea invisibly by romanizing queries at input time rather than asking users to type in a particular script, which would make the script gap a deployment detail rather than a user burden.
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

4 major / 5 minor

Summary. The paper studies how multilingual neural IR systems behave when queries are transliterated from a non-Latin script (Chinese, Russian) into Latin script. Using Uroman as the romanization tool, the authors build transliterated versions of mMARCO/v2 and NeuCLIR/1 queries and fine-tune BGE-M3 and an mT5-based reranker under three training configurations: native only (N), an even native/transliterated mixture (50), and transliterated only (T). They report that off-the-shelf models degrade sharply on transliterated queries, that the mixed configuration substantially improves transliterated-query effectiveness in-domain while retaining native-query performance for Russian, that the transliterated-only configuration behaves differently across languages and models, and that out-of-domain evaluation on NeuCLIR shows gains for Russian but not for Chinese. The paper concludes that a 'transliterate-train' procedure can bridge the script gap on the query side.

Significance. The proposed recipe is simple and potentially useful: adding transliterated training queries to the fine-tuning mixture improves robustness to Uroman-style romanized queries in both tested languages, with Russian reaching near parity and Chinese showing large in-domain gains. The evaluation is on standard held-out benchmarks (mMARCO dev, TREC NeuCLIR), covers two model families (dense retriever and cross-encoder reranker), and uses significance testing with Bonferroni correction. There are no fitted parameters or circular derivations: the method is an augmentation strategy evaluated on held-out data. The main significance caveat is that the evidence is limited to Uroman's deterministic romanizations, so the headline claim about bridging the script gap for real users' transliterations is not yet established. The paper is a reasonable short-paper contribution if the claims are scoped to Uroman-style transliterations and if the abstract is aligned with the Chinese residual gap.

major comments (4)
  1. [Abstract / §4.2, Table 1] The abstract's claim that models 'can perform this cross-script matching at nearly the same performance as when the query was formulated in the native script' is only true for Russian. For Chinese under the mixed configuration, Table 1 shows MRR@10 of 0.1382 for transliterated queries versus 0.2642 for native queries with BGE-M3, and 0.1105 versus 0.2520 with the mT5 cascade; a residual gap of roughly half remains. The conclusion in §4.2 that 'we can bridge the gap between the transliterated and native scripts' should therefore be qualified to say that the gap is fully bridged for Russian and only partially bridged for Chinese.
  2. [§4.2, Table 1] The statement that 'the mixed configuration is the only fine-tuning configuration which preserves the native text performance and significantly enhances the robustness of both models on the transliterated queries' is contradicted by the Russian rows of Table 1. For BGE-M3, the transliterated-only (T) configuration preserves native MRR@10 (0.2785 vs. the 0.2444 baseline) and improves transliterated MRR@10 from 0.1244 to 0.2679; for mT5, the T configuration also preserves native MRR@10 (0.2662 vs. 0.2552) and improves transliterated MRR@10 from 0.1948 to 0.2522. The uniqueness claim is only true for Chinese; the conclusion should be rewritten as a language-dependent finding.
  3. [§3 / §4] All training and evaluation transliterations are generated by Uroman, a deterministic romanizer, yet the introduction motivates the problem with heterogeneous real-world transliterations such as Greeklish and Arabizi, which contain vowel drops, digit substitutions, and non-canonical spellings. Because the training and evaluation data share the same transform, the reported gains may not transfer to the transliterations real users actually type. The authors should either evaluate on real transliteration data (for example, FIRE transliterated-search queries or Greeklish samples) or explicitly scope the contribution to Uroman-style romanizations and state this as a limitation.
  4. [§4.1 / §4.2] The manuscript reports that pre-tokenizing Chinese queries before transliteration yields inconsistent results between BGE-M3 and mT5 and then excludes this variant from the main experiments. Since Uroman's Chinese output appears to have no word boundaries and the residual Chinese script gap is large, the tokenization choice could materially affect the central comparison. The authors should report the pre-tokenization results (even in an appendix or the linked repository) or provide a more direct justification that the choice is neutral for the claims.
minor comments (5)
  1. [§4] The paper does not report fine-tuning hyperparameters (epochs, learning rate, batch size, number of training queries, random seeds, or number of runs). Since the contribution is a training recipe, these details are needed for reproducibility; they should be added to the text or referenced to a specific committed version of the GitHub repository.
  2. [Tables 1 and 2] The significance testing is described only as 'Bonferroni correction' with a dagger denoting significance. Please specify the underlying test (e.g., paired t-test or permutation test) and clarify whether the dagger marks significant improvement, significant difference in either direction, or significance after correction; this is currently ambiguous in cells where the metric is lower than the baseline.
  3. [§4.3] The qualitative analysis is based on ten sampled queries and informal overlap counts. It is useful as illustration, but the paper should state explicitly that this is not a systematic evaluation and should avoid drawing strong conclusions from the number of common documents across scripts.
  4. [§2] The FIRE transliterated-search and mixed-script tasks are cited in related work but not used for evaluation. A sentence explaining why these collections were not suitable (for example, different languages or BM25-era evaluation practices) would help readers understand the experimental scope.
  5. [§4.3] The Chinese query examples in the submitted text appear as broken glyphs (e.g., the strings around 'qid 32' and 'qid 26'). Please ensure the final PDF renders the native-script examples correctly.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central claim is supported by held-out mMARCO dev and out-of-domain NeuCLIR evaluations, without fitted parameters or load-bearing self-citations.

full rationale

The paper's central claim (RQ2, §4.2) is an empirical finding: fine-tuning BGE-M3 and mT5 on an even mixture of native and Uroman-transliterated queries improves transliterated query effectiveness while preserving native effectiveness. This is established by comparing three fine-tuning configurations (N, 50, T) on the mMARCO/v2 dev set and by out-of-domain evaluation on TREC NeuCLIR 2022/2023 (§4.3). No parameter is fitted to the evaluation data, and no prediction is derived from an input that already contains the output; the "transliterate-train" procedure (§3) is a data augmentation step whose effect is measured independently on held-out queries. The only self-citation, [4] (Chari et al., SIGIR 2023), is cited in §2 as motivation for studying surface-form variation and is explicitly described as limited to British/American spelling; it does not supply any premise of the present argument. The paper also reports mixed out-of-domain results (e.g., Chinese transliterated nDCG@20 remains low in Table 2) and a qualitative analysis of lost nuance, which further shows the conclusions are not forced. The use of Uroman for both training and evaluation is a scope limitation about the realism of transliteration, not circularity: the method is tested on the same canonical romanization it is trained on, which is a valid in-domain evaluation of the proposed fine-tuning strategy, and the paper does not claim to have modelled ad hoc user transliterations.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

No free parameters are fit to the evaluation data; the method is standard fine-tuning on existing labelled data (mMARCO) with three fixed mixture ratios (N, 50, T). No new entities are postulated; only a training data augmentation strategy. The key unverified assumptions are the representativeness of Uroman transliterations and of the automatically translated mMARCO labels.

assumptions (3)
  • domain assumption mMARCO Chinese/Russian relevance labels are reliable enough to support fine-tuning and evaluation.
    The paper's claims rest on the assumption that the automatically translated mMARCO resource is a valid surrogate for real retrieval relevance; the authors rely on this without validation beyond prior literature.
  • domain assumption Uroman's romanizations are representative of real users' transliterated queries.
    The entire transliterate-train approach assumes that canonical Uroman transliterations approximate how users type (e.g., Greeklish or Arabizi). Real user transliterations are often noisier, with dropped vowels or digit substitutions, which the paper does not test.
  • domain assumption BGE-M3 single-vector representations and mT5 reranking are representative of current multilingual neural IR.
    The paper generalizes from these two models to 'multilingual IR methods', so the representativeness of the chosen models is an implicit assumption about the broader model family.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Lost in Transliteration: Bridging the Script Gap in Neural IR." pith.science (2026). https://pith.science/paper/MTEU3CAJ

@misc{pith2026250508411,
  author       = {Pith},
  title        = {Pith review of: Lost in Transliteration: Bridging the Script Gap in Neural IR},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MTEU3CAJ}},
  note         = {Machine review of arXiv:2505.08411}
}
read the original abstract

Most human languages use scripts other than the Latin alphabet. Search users in these languages often formulate their information needs in a transliterated -- usually Latinized -- form for ease of typing. For example, Greek speakers might use Greeklish, and Arabic speakers might use Arabizi. This paper shows that current search systems, including those that use multilingual dense embeddings such as BGE-M3, do not generalise to this setting, and their performance rapidly deteriorates when exposed to transliterated queries. This creates a ``script gap" between the performance of the same queries when written in their native or transliterated form. We explore whether adapting the popular ``translate-train" paradigm to transliterations can enhance the robustness of multilingual Information Retrieval (IR) methods and bridge the gap between native and transliterated scripts. By exploring various combinations of non-Latin and Latinized query text for training, we investigate whether we can enhance the capacity of existing neural retrieval techniques and enable them to apply to this important setting. We show that by further fine-tuning IR models on an even mixture of native and Latinized text, they can perform this cross-script matching at nearly the same performance as when the query was formulated in the native script. Out-of-domain evaluation and further qualitative analysis show that transliterations can also cause queries to lose some of their nuances, motivating further research in this direction.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 14 canonical work pages

  1. [4]

    Andreas Chari, Sean MacAvaney, and Iadh Ounis. 2023. On the Effects of Re- gional Spelling Conventions in Retrieval Models. InProceedings of the 46th In- ternational ACM SIGIR Conference on Research and Development in Information Retrieval (Taipei, Taiwan) (SIGIR ’23). Association for Computing Machinery, New York, NY, USA, 2220⚶2224. doi:10.1145/3539618.3592030

    arXiv 2023

  2. [1]

    Salman Al-Azami, Charmian Kenner, Mahera Ruby, and Eve Gregory. 2010. Transliteration as a bridge to learning for bilingual children.International Jour- nal of Bilingual Education and Bilingualism 13, 6 (2010), 683–700

    work page 2010

  3. [2]

    Somnath Banerjee, Kunal Chakma, Sudip Kumar Naskar, Amitava Das, Paolo Rosso, Sivaji Bandyopadhyay, and Monojit Choudhury. 2016. Overview of the MixedScriptInformationRetrieval(MSIR)atFIRE-2016.In Working notes of FIRE 2016 - Forum for Information Retrieval Evaluation, Kolkata, India, December 7- 10, 2016 (CEUR Workshop Proceedings, Vol. 1737) , Prasenjit...

    work page 2016

  4. [3]

    Luiz Henrique Bonifacio, Israel Campiotti, Roberto de Alencar Lotufo, and Rodrigo Frassetto Nogueira. 2021. mMARCO: A Multilingual Version of MS MARCO Passage Ranking Dataset. CoRR abs/2108.13897 (2021). arXiv:2108.13897 https://arxiv.org/abs/2108.13897

    arXiv 2021

  5. [5]

    Jianlv Chen, Shitao Xiao, Peitian Zhang, Kun Luo, Defu Lian, and Zheng Liu. 2024. BGE M3-Embedding: Multi-Lingual, Multi-Functionality, Multi- Granularity Text Embeddings Through Self-Knowledge Distillation. CoRR abs/2402.03216 (2024).doi:10.48550/ARXIV.2402.03216 arXiv:2402.03216

    arXiv doi:10.48550/arxiv.2402.03216 2024

  6. [6]

    Monojit Choudhury, Gokul Chittaranjan, Parth Gupta, and Amitava Das. 2014. Overview of fire 2014 track on transliterated search.Proceedings of FIRE (2014), 68–89

    work page 2014

  7. [7]

    Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guil- laume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer, and Veselin Stoyanov. 2020. Unsupervised Cross-lingual Representation Learn- ing at Scale. InProceedings of the 58th Annual Meeting of the Association for Com- putational Linguistics, ACL 2020, Online, July ...

    2020

  8. [8]

    Brenda Danet and Susan C. Herring. 2007.The Multilingual Internet: Language, Culture, and Communication Online . Oxford University Press. doi:10.1093/ acprof:oso/9780195304794.001.0001

    arXiv 2007

Show all 30 references
  1. [9]

    Ulf Hermjakob, Jonathan May, and Kevin Knight. 2018. Out-of-the-box Univer- salRomanizationTooluroman.In Proceedings of ACL 2018, Melbourne, Australia, July 15-20, 2018, System Demonstrations , Fei Liu and Thamar Solorio (Eds.). As- sociation for Computational Linguistics, 13–...

    2018 doi

  2. [10]

    Jaavid Aktar Husain, Raj Dabre, Aswanth M., Jay Gala, Thanmay Jayakumar, Ratish Puduppully, and Anoop Kunchukuttan. 2024. RomanSetu: Efficiently un- locking multilingual capabilities of Large Language Models via Romanization. In Proceedings of the 62nd Annual Meeting of the As...

    2024

  3. [11]

    Lawrie, Sean MacAvaney, James Mayfield, Paul McNamee, Douglas W

    Dawn J. Lawrie, Sean MacAvaney, James Mayfield, Paul McNamee, Douglas W. Oard, Luca Soldaini, and Eugene Yang. 2023. Overview of the TREC 2022 NeuCLIR Track. CoRR abs/2304.12367 (2023). doi:10.48550/ARXIV.2304.12367 arXiv:2304.12367

    2023 doi arXiv

  4. [12]

    Lawrie, Sean MacAvaney, James Mayfield, Paul McNamee, Douglas W

    Dawn J. Lawrie, Sean MacAvaney, James Mayfield, Paul McNamee, Douglas W. Oard, Luca Soldaini, and Eugene Yang. 2024. Overview of the TREC 2023 NeuCLIR Track. CoRR abs/2404.08071 (2024). doi:10.48550/ARXIV.2404.08071 arXiv:2404.08071

    2024 doi arXiv

  5. [13]

    Lawrie, Eugene Yang, Douglas W

    Dawn J. Lawrie, Eugene Yang, Douglas W. Oard, and James Mayfield. 2023. Neu- ralApproachestoMultilingualInformationRetrieval.In Advances in Information Retrieval - 45th European Conference on Information Retrieval, ECIR 2023, Dublin, Ireland, April 2-6, 2023, Proceedings, Part...

    2023 doi

  6. [14]

    Yihong Liu, Chunlan Ma, Haotian Ye, and Hinrich Schütze. 2024. TransliCo: A Contrastive Learning Framework to Address the Script Barrier in Multilin- gual Pretrained Language Models. InProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Vol...

    2024 doi

  7. [15]

    Chunlan Ma, Yihong Liu, Haotian Ye, and Hinrich Schütze. 2024. Exploring the Role of Transliteration in In-Context Learning for Low-resource Languages Written in Non-Latin Scripts.CoRR abs/2407.02320 (2024).doi:10.48550/ARXIV. 2407.02320 arXiv:2407.02320

    2024 doi arXiv

  8. [16]

    Sean MacAvaney, Craig Macdonald, and Iadh Ounis. 2022. Streamlining Eval- uation with ir-measures. InAdvances in Information Retrieval - 44th European Conference on IR Research, ECIR 2022, Stavanger, Norway, April 10-14, 2022, Pro- ceedings, Part II (Lecture Notes in Computer ...

    2022 doi

  9. [17]

    SeanMacAvaney,AndrewYates,SergeyFeldman,DougDowney,ArmanCohan, and Nazli Goharian. 2021. Simplified Data Wrangling with ir_datasets. InSIGIR ’21: The 44th International ACM SIGIR Conference on Research and Development in Information Retrieval, Virtual Event, Canada, July 11-15...

    2021

  10. [18]

    Craig Macdonald and Nicola Tonellotto. 2020. Declarative Experimentation in Information Retrieval using PyTerrier. InICTIR ’20: The 2020 ACM SIGIR Interna- tional Conference on the Theory of Information Retrieval, Virtual Event, Norway, September 14-17, 2020, Krisztian Balog, ...

    2020 doi

  11. [19]

    Mansurov and A

    B. Mansurov and A. Mansurov. 2021. UzBERT: pretraining a BERT model for Uzbek. CoRR abs/2108.09814 (2021). arXiv: 2108.09814 https://arxiv.org/abs/ 2108.09814

    2021 arXiv

  12. [20]

    Lawrie, Kevin Duh, Paul McNamee, Kenton Murray, James Mayfield, and Douglas W

    Suraj Nair, Eugene Yang, Dawn J. Lawrie, Kevin Duh, Paul McNamee, Kenton Murray, James Mayfield, and Douglas W. Oard. 2022. Transfer Learning Ap- proaches for Building Cross-Language Dense Retrieval Models. InAdvances in Information Retrieval - 44th European Conference on IR R...

    2022

  13. [21]

    Tri Nguyen, Mir Rosenberg, Xia Song, Jianfeng Gao, Saurabh Tiwary, Ran- gan Majumder, and Li Deng. 2016. MS MARCO: A Human Generated MA- chine Reading COmprehension Dataset. InProceedings of the Workshop on Cog- nitive Computation: Integrating neural and symbolic approaches 20...

    2016

  14. [22]

    oaFunkyLanguageforTeenzz to Use:pRepresenting Gulf Arabic in Instant Messaging.Journal of Computer- Mediated Communication 9, 1 (07 2017), JCMC917.doi:10.1111/j.1083-6101.2003

    DavidPalfreymanandMuhamedalKhalil.2017. oaFunkyLanguageforTeenzz to Use:pRepresenting Gulf Arabic in Instant Messaging.Journal of Computer- Mediated Communication 9, 1 (07 2017), JCMC917.doi:10.1111/j.1083-6101.2003. tb00355.x

    2017

  15. [23]

    Dinesh Kumar Prabhakar and Sukomal Pal. 2018. Machine transliteration and transliterated text retrieval: a survey. Sādhanā 43, 6 (07 Jun 2018), 93.doi:10. 1007/s12046-018-0828-8

    2018

  16. [24]

    ExploringtheLimits of Transfer Learning with a Unified Text-to-Text Transformer.J

    Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, MichaelMatena, YanqiZhou, WeiLi, andPeterJ.Liu.2020. ExploringtheLimits of Transfer Learning with a Unified Text-to-Text Transformer.J. Mach. Learn. Res. 21 (2020), 140:1–140:67. https://jmlr.org/papers/v2...

    2020

  17. [25]

    Rishiraj Saha Roy, Monojit Choudhury, Prasenjit Majumder, and Komal Agar- wal. 2013. Overview of the FIRE 2013 Track on Transliterated Search. InPro- ceedings of the 5th 2013 Forum on Information Retrieval Evaluation, FIRE ’13, New Delhi, India, December 4-6, 2013 , Prasenjit ...

    2013

  18. [26]

    Royal Sequiera, Monojit Choudhury, Parth Gupta, Paolo Rosso, Shubham Ku- mar, Somnath Banerjee, Sudip Kumar Naskar, Sivaji Bandyopadhyay, Gokul Chittaranjan, Amitava Das, and Kunal Chakma. 2015. Overview of FIRE-2015 Shared Task on Mixed Script Information Retrieval. InPost Pr...

    2015

  19. [27]

    Anastasios Toumazatos, John Pavlopoulos, Ion Androutsopoulos, and Stavros Vassos. 2024. Still All Greeklish to Me: Greeklish to Greek Transliteration. In Proceedings of the 2024 Joint International Conference on Computational Linguis- tics, Language Resources and Evaluation, L...

    2024

  20. [28]

    Orgest Xhelili, Yihong Liu, and Hinrich Schütze. 2024. Breaking the Script Bar- rier in Multilingual Pre-Trained Language Models with Transliteration-Based Post-Training Alignment. InFindings of the Association for Computational Lin- guistics: EMNLP 2024, Miami, Florida, USA, ...

    2024

  21. [29]

    Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, and Colin Raffel. 2021. mT5: A Massively Multilin- gual Pre-trained Text-to-Text Transformer. InProceedings of the 2021 Confer- ence of the North American Chapter of the Associat...

    2021

  22. [30]

    Mahima Yadav, Ishan Kumar, and Ayush Kumar. 2023. Different Models of Transliteration - A Comprehensive Review. In2023 International Conference on Innovative Data Communication Technologies and Application (ICIDCA).356–363. doi:10.1109/ICIDCA56705.2023.10099632

    2023

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.