Pith. sign in

REVIEW 4 major objections 5 minor 34 references

Multi-label Scandinavian Language Identification (SLIDE)

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

Pith's one-line read SLIDE demonstrates that accurate language identification among Danish, Norwegian Bokmål, Nynorsk, and Swedish requires multi-label output, and its best model reaches 96.4% exact-match accuracy on a manually annotated test set.

desk verdict A useful dataset and models for multi-label Scandinavian LID, but the necessity claim exceeds the evidence and the annotation quality is under-documented. read the letter →

arxiv 2502.06692 v1 pith:ISSQPWTA submitted 2025-02-10 cs.CL cs.AI

classification cs.CLcs.AI
keywords languageidentificationmulti-labelclassificationScandinavianlanguagesNorwegianBokmålNynorskcloselyrelatedsilverlabelingmachinetranslation
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

SLIDE argues that sentence-level language identification (LID) among the four Mainland Scandinavian languages cannot be reduced to picking one label: a meaningful share of sentences—about 5% of the human-annotated test set and 16% of sentences shorter than six words—are valid in more than one of Danish, Bokmål, Nynorsk, and Swedish. Because of this, the paper claims, any accurate LID method must be able to output several languages at once, and single-label evaluation masks the failures. To train such a system without expensive manual labels, the authors convert machine translation into a labeling step: a sentence gets an extra language label whenever a translator leaves it unchanged. They release the resulting SLIDE dataset and models; the best model reaches 96.4% exact-match accuracy on the new test set, above all evaluated single-label baselines.

What carries the argument

The load-bearing mechanism is the translation-identity heuristic. The authors start from single-label sentences from existing treebank data, then translate each sentence into each of the other three target languages; whenever the translator returns the sentence unchanged, they add that language as a silver label. The heuristic relies on translation models' observed conservatism—they minimize edits—so an unchanged output is read as evidence that the source sentence is already acceptable in the target language. Training then proceeds as multi-label classification with binary cross-entropy on masked-language models pretrained for Norwegian (the SLIDE-xs/small/base models) and on an embedding-based model (SLIDE-fast). Two evaluation metrics carry the comparison: loose accuracy, which forgives a single-label model for choosing any gold label, and exact-match accuracy, which requires the full predicted label set to equal the gold set.

What would settle it

Randomly sample sentences from Scandinavian text, have native speakers of each of the four languages mark every language in which each sentence is acceptable, and compare the multi-valid share with the paper's 5% figure; if a large representative sample shows the share is near zero, the claim that accurate LID requires multi-label output loses its empirical footing.

Watch

Extended reading notes

Core claim

The paper's central discovery is empirical and methodological at once. Empirically, manually re-examining treebank sentences shows that multi-label instances are not rare noise: they make up about 5% of the test set, and short sentences are disproportionately affected (16% under six words). Methodologically, the paper shows that a machine translation model can be used as a conservative annotator—if the translation of a sentence into another Scandinavian language changes nothing, the sentence is labeled as valid in that language—which lets a single-label corpus be extended into a multi-label training set at no manual cost. On the resulting SLIDE test set, the best multi-label model attains 96.4% exact-match accuracy, outperforming all evaluated baselines including the best single-label model, and the fast embedding-based variant matches the best baseline's exact-match accuracy while improving F1 on three of the four languages. The paper takes these results to show both that multi-label output is necessary for accurate LID among closely related languages and that its translation-based labeling method is sufficient to train such models.

Load-bearing premise

The load-bearing assumption is that a machine translation model that leaves a sentence unchanged when translating it into another language reliably means the sentence is acceptable in that language; if translators are too conservative, the automatically generated multi-label training data is silently corrupted.

Editorial extensions

If this is right

  • LID benchmarks for closely related languages should use multi-label gold sets and exact-match accuracy; loose accuracy alone hides exactly the ambiguity the paper documents.
  • Off-the-shelf single-label LID tools will systematically mishandle the roughly 5% of multi-valid sentences, and much more so for short texts, where multi-validity reaches 16%.
  • The translation-identity labeling method can be reused for other closely related language groups whenever a reliable translator between those languages exists.
  • The SLIDE model family gives practitioners a clear speed-accuracy spectrum: the fast model suits high-throughput preprocessing, while the base model maximizes accuracy at about 96.4% exact-match accuracy.

Reading between the lines

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

  • The identity heuristic's reliability could be measured directly: sample silver-labeled sentences where the translator output was unchanged and ask native speakers whether the sentence is truly acceptable in the target language; that would give a noise estimate the paper does not provide.
  • The same multi-label framing should transfer to other tight language clusters such as Bosnian, Croatian, Montenegrin and Serbian, or Arabic dialects; the paper's necessity argument is not Scandinavia-specific, though the authors do not test these cases.
  • Because the training data is dominated by Bokmål and the annotation was done by Norwegian speakers, the consistently lower Danish F1 could partly reflect data skew and annotator asymmetry rather than an intrinsic property of Danish; a balanced Danish addition would be a cheap check.
  • The decision to ignore punctuation differences when labeling Danish and Bokmål ambiguity is a policy choice; systems for downstream tasks that care about orthographic correctness would need different labels.
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.

Referee Report

4 major / 5 minor

Summary. The paper introduces SLIDE, a manually annotated multi-label evaluation dataset for sentence-level language identification among Danish, Norwegian Bokmål, Nynorsk, and Swedish, built from Universal Dependencies treebank splits. The authors also release a family of LID models—three NorBERT3-based transformers and a fast static-embedding model—trained with a proposed silver-labeling method that uses machine translation identity: if a sentence is unchanged by translation into a target language, it is labeled as valid in that language. On the SLIDE test set, the best model reaches 96.4% exact-match accuracy, outperforming baselines such as GlotLID (93.4%), and the paper claims that multi-label output is necessary for any accurate LID method. The authors additionally report out-of-domain results on the Haas and Derczynski (2021) test sets, where SLIDE-base is slightly below GlotLID.

Significance. The manually curated multi-label dataset is a useful resource, and the released models with reported speed-accuracy trade-offs are a practical contribution. The evaluation is not circular: the test set labels are manually annotated and independent of the training pipeline, and the MT-identity heuristic produces training labels rather than model-derived predictions. However, the central abstract claim is considerably stronger than the evidence: it rests on one test set, the annotation reliability of that test set is not demonstrated, the MT silver-labeling heuristic is never validated, and the out-of-domain results do not support the word 'any.' With targeted additional analyses and a more careful claim, the contributions would be publishable; as written, the central claim overreaches.

major comments (4)
  1. [Abstract and §6.2] The claim that multi-label output is 'necessary for any accurate LID method' is not supported by the reported experiments. Table 4 shows a three-point exact-match gap between SLIDE-base (96.4) and GlotLID (93.4) on a single 6,950-sentence test set, but exact-match accuracy by construction gives no credit to single-label classifiers on the 5% of gold instances that are multi-label, so the gap conflates multi-label handling with overall classifier quality. To make the claim load-bearing, please report performance on the single-label subset, give an upper bound for any single-label classifier on the multi-label instances (for example, loose accuracy restricted to those instances), and quantify how much of the exact-match gap is actually attributable to multi-label gold instances rather than ordinary misclassification. Absent that analysis, the claim should be narrowed to 'on the SLIDE test set, exact-match accuracy requires multi-label predictions.'
  2. [§3.3] The MT-identity silver-labeling heuristic is never validated. The method assumes that NorMistral is conservative and that an unchanged translation is a reliable positive signal for the target language, but no evidence is provided for this assumption. Since the multi-label training signal and the claimed 'novel approach' both depend on this heuristic, please validate it directly, for example by comparing MT-derived multi-labels against the manually annotated SLIDE validation/test instances, or by manual inspection of a sample of identity translations. Without such validation, the reported multi-label model performance could reflect noise from erroneous silver labels rather than genuine multi-label supervision.
  3. [§3.2 and Limitations] The manual test set, which is the sole empirical basis for the central claim, lacks inter-annotator agreement statistics, and the Limitations section concedes that the Norwegian annotators' productive skills in Danish and Swedish are lower and that mislabeling may occur. The annotation conventions for proper-name-only sentences and for optional comma placement also directly increase the multi-label rate: proper-name-only instances are labeled with all four Scandinavian languages, and Danish/Bokmål sentences differing only by a comma are labeled as both. Please report agreement scores, provide a breakdown of multi-label instances by category, and run a sensitivity analysis excluding or re-labeling the contested categories. If the 5% multi-label rate changes substantially under that analysis, the necessity claim is weakened.
  4. [§6.3] The out-of-domain results weaken the generality of the 'any accurate LID method' claim. On the Haas and Derczynski (2021) test sets, SLIDE-base achieves 92.7 and 95.3 accuracy, slightly below GlotLID's 93.0 and 95.7. The authors attribute this to preprocessing and mislabeled sentences, but those are exactly the conditions under which a generally accurate LID method must operate. Either temper the abstract claim to the SLIDE test set or present additional evidence, such as results on other out-of-domain data, that SLIDE retains its advantage beyond the treebank-derived evaluation set.
minor comments (5)
  1. [§3.2] The sentence 'We also find 29444 examples of a comma preceding at 'that' in the Norwegian LBK corpus' appears to contain a typo and unclear phrasing; it should read something like '29,444 examples of a comma preceding 'at'' or 'a comma before the conjunction 'that''.
  2. [Table 1] The caption states that multi-label samples are reported once for each language, but the table does not show the number of multi-label instances; please clarify how the 5% and 6% multi-label rates are computed from the table, or add a multi-label row or column.
  3. [§5.2] The sigmoid threshold of 0.5 is selected based on the validation split; this is a free hyperparameter, and since SLIDE-fast's exact-match accuracy depends on it, please report its sensitivity or justify the choice with a small threshold sweep.
  4. [References] The HeLI-OTS reference appears twice as Jauhiainen et al. (2022a) and (2022b) with identical titles and page ranges; these appear to be duplicate entries and should be consolidated.
  5. [Table 3] The caption note 'NorBERT3 refers to the same model as SLIDE' is confusing because the table is about base-model selection; please clarify that the final model is named SLIDE-base.

Circularity Check

0 steps flagged · score 1.0 of 10

No circular derivation: the SLIDE models are evaluated on an independently constructed manual test set, and self-citations point to released checkpoints rather than to the target result.

full rationale

The paper's central empirical chain is training (Sections 3.3 and 5) followed by evaluation on the manually annotated SLIDE test set (Sections 3.2 and 6.2); no test-set labels or parameters are fitted into the models, and the best checkpoints and the SLIDE-fast threshold are selected on validation data, so no fitted input is renamed as a prediction. The MT-identity silver labeling in Section 3.3 is a training-data heuristic, not a derived result, and its output is checked against the independent manual test set. The only self-citations, to NorBERT3 and NorMistral, are to externally released pretrained models used as backbones and as a translation tool; they do not import the paper's own conclusion. The Limitations paragraph does concede that Norwegian annotators' lower productive skills in Danish and Swedish 'might be cases of mislabeling,' and the annotation conventions for proper names, numbers, and optional commas inflate the measured multi-label rate; however, those are disclosed data-quality and evaluation-design limitations rather than circular reductions. Under the exact-match metric, multi-label output is necessary for perfect scores by definition, but the paper's test-set contribution is independently constructed and its model rankings are not forced by that definition alone.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical or conceptual entities; the SLIDE dataset is a curated collection, not an invented entity in the sense of a new particle or force. The key free parameters are model and augmentation hyperparameters, and the main assumptions are about the quality of source labels and the MT identity heuristic.

free parameters (3)
  • SLIDE-fast sigmoid threshold = 0.5
    Selected based on the validation data split (Section 5.2). This threshold decides which labels are accepted and affects the multi-label output of the fast model.
  • Punctuation augmentation proportion = 7.5% of training data
    Heuristically chosen (Section 3.4). The amount of punctuation perturbation is not derived from data and could influence model robustness.
  • Intervening space probability in punctuation augmentation = 1/3
    Chosen as part of the augmentation scheme in Section 3.4. This probability is arbitrary and affects the augmented training distribution.
assumptions (4)
  • domain assumption UD treebank language tags are accurate single-language labels for each sentence.
    Section 3.1 states 'we assume that the texts accurately reflect their corresponding languages'. The entire dataset construction builds on this.
  • domain assumption Machine translation identity implies sentence validity in the target language.
    Section 3.3: 'if the translation of a sentence does not lead to any changes, we label it as a valid sentence of the target language'. The silver-labeling method depends on this assumption, which is never directly validated.
  • domain assumption Native or near-native Norwegian speakers can reliably judge Danish and Swedish acceptability.
    Section 3.2 describes annotators' exposure to other Scandinavian languages through education, but the Limitations section acknowledges possible mislabeling due to lower productive proficiency in Danish and Swedish.
  • ad hoc to paper Proper names and numbers are valid in all Scandinavian languages.
    Section 3.2: 'Instances consisting of only proper names were annotated with all Scandinavian languages'. This annotation policy directly shapes the multi-label gold labels.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multi-label Scandinavian Language Identification (SLIDE)." pith.science (2026). https://pith.science/paper/ISSQPWTA

@misc{pith2026250206692,
  author       = {Pith},
  title        = {Pith review of: Multi-label Scandinavian Language Identification (SLIDE)},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ISSQPWTA}},
  note         = {Machine review of arXiv:2502.06692}
}
read the original abstract

Identifying closely related languages at sentence level is difficult, in particular because it is often impossible to assign a sentence to a single language. In this paper, we focus on multi-label sentence-level Scandinavian language identification (LID) for Danish, Norwegian Bokm\r{a}l, Norwegian Nynorsk, and Swedish. We present the Scandinavian Language Identification and Evaluation, SLIDE, a manually curated multi-label evaluation dataset and a suite of LID models with varying speed-accuracy tradeoffs. We demonstrate that the ability to identify multiple languages simultaneously is necessary for any accurate LID method, and present a novel approach to training such multi-label LID models.

Figures

Figures reproduced from arXiv: 2502.06692 by the authors.

Figure 1
Figure 1. Scandinavian similarity Accurate lan￾guage identification has to necessarily be multi￾label when discriminating between closely related languages. In this paper, we focus on the four closely re￾lated Scandinavian languages: Danish, Norwegian Bokmal, Norwegian Nynorsk, and Swedish. In or- ˚ der to accurately differentiate within this group, we move away from the standard single-label (multi￾class) language identifica… view at source ↗
Figure 2
Figure 2. Training pipeline A diagram that il￾lustrates the flow of the full training pipeline. We start with a high-quality, single-labeled training dataset, then extend it with multi-label annotations using a strong machine translation model. The dataset is further augmented by randomly swap￾ping named entities identified by existing NER models and through other rule-based augmenta￾tions. We use the (augmented) data to fine… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 17 canonical work pages

  1. [1]

    URL: " 'urlintro :=

    ENTRY address author booktitle chapter edition editor howpublished institution journal key month note number organization pages publisher school series title type volume year eprint doi pubmed url lastchecked label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block STRINGS urlintro eprinturl eprintpr...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    Muhammad Abdul-Mageed, Amr Keleg, AbdelRahim Elmadany, Chiyu Zhang, Injy Hamed, Walid Magdy, Houda Bouamor, and Nizar Habash. 2024. https://doi.org/10.18653/v1/2024.arabicnlp-1.79 NADI 2024: The fifth nuanced A rabic dialect identification shared task . In Proceedings of The Second Arabic Natural Language Processing Conference, pages 709--728, Bangkok, Th...

  4. [4]

    Jason Ansel, Edward Yang, Horace He, Natalia Gimelshein, Animesh Jain, Michael Voznesensky, Bin Bao, Peter Bell, David Berard, Evgeni Burovski, Geeta Chauhan, Anjali Chourdia, Will Constable, Alban Desmaison, Zachary DeVito, Elias Ellison, Will Feng, Jiong Gong, Michael Gschwind, Brian Hirsh, Sherlock Huang, Kshiteej Kalambarkar, Laurent Kirsch, Michael L...

  5. [5]

    Laurie Burchell, Alexandra Birch, Nikolay Bogoychev, and Kenneth Heafield. 2023. https://doi.org/10.18653/v1/2023.acl-short.75 An open dataset and model for language identification . In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 865--879, Toronto, Canada. Association for Computat...

  6. [6]

    Adrian-Gabriel Chifu, Goran Glava s , Radu Tudor Ionescu, Nikola Ljube s i \'c , Aleksandra Mileti \'c , Filip Mileti \'c , Yves Scherrer, and Ivan Vuli \'c . 2024. https://doi.org/10.18653/v1/2024.vardial-1.1 V ar D ial evaluation campaign 2024: Commonsense reasoning in dialects and multi-label similar language identification . In Proceedings of the Elev...

  7. [7]

    Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzm \'a n, Edouard Grave, Myle Ott, Luke Zettlemoyer, and Veselin Stoyanov. 2020. https://doi.org/10.18653/v1/2020.acl-main.747 Unsupervised cross-lingual representation learning at scale . In Proceedings of the 58th Annual Meeting of the Association for Comp...

  8. [8]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. https://doi.org/10.18653/v1/N19-1423 BERT : Pre-training of deep bidirectional transformers for language understanding . In Proceedings of the 2019 Conference of the North A merican Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long a...

Show all 34 references
  1. [9]

    Ruth Vatvedt Fjeld, Anders Nøklestad, and Kristin Hagen. 2020. Leksikografisk bokmålskorpus ( LBK ) – bakgrunn og bruk. Oslo Studies in Language, 11(1):47--59

  2. [10]

    Cyril Goutte, Serge L \'e ger, Shervin Malmasi, and Marcos Zampieri. 2016. https://aclanthology.org/L16-1284 Discriminating similar languages: Evaluations and explorations . In Proceedings of the Tenth International Conference on Language Resources and Evaluation ( LREC '16) ,...

  3. [11]

    Edouard Grave, Piotr Bojanowski, Prakhar Gupta, Armand Joulin, and Tomas Mikolov. 2018. https://aclanthology.org/L18-1550 Learning word vectors for 157 languages . In Proceedings of the Eleventh International Conference on Language Resources and Evaluation ( LREC 2018) , Miyaz...

  4. [12]

    Ren \'e Haas and Leon Derczynski. 2021. https://aclanthology.org/2021.vardial-1.8 Discriminating between similar N ordic languages . In Proceedings of the Eighth Workshop on NLP for Similar Languages, Varieties and Dialects, pages 67--75, Kiyv, Ukraine. Association for Computa...

  5. [13]

    Tommi Jauhiainen, Heidi Jauhiainen, and Krister Lind \'e n. 2022 a . http://www.lrec-conf.org/proceedings/lrec2022/pdf/2022.lrec-1.416.pdf H e LI-OTS , off-the-shelf language identifier for text . In Proceedings of the 13th Conference on Language Resources and Evaluation, page...

  6. [14]

    Tommi Jauhiainen, Heidi Jauhiainen, and Krister Lind \'e n. 2022 b . https://aclanthology.org/2022.lrec-1.416 H e LI - OTS , off-the-shelf language identifier for text . In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 3912--3922, Marseille,...

  7. [15]

    Tommi Jauhiainen, Marco Lui, Marcos Zampieri, Timothy Baldwin, and Krister Lind \'e n. 2019. https://doi.org/10.1613/jair.1.11675 Automatic language identification in texts: A survey . Journal of Artificial Intelligence Research, 65:675--782

  8. [16]

    Armand Joulin, Edouard Grave, Piotr Bojanowski, and Tomas Mikolov. 2017. https://aclanthology.org/E17-2068 Bag of tricks for efficient text classification . In Proceedings of the 15th Conference of the E uropean Chapter of the Association for Computational Linguistics: Volume ...

  9. [17]

    Amir Hossein Kargaran, Ayyoob Imani, Fran c ois Yvon, and Hinrich Sch \"u tze. 2023. https://openreview.net/forum?id=dl4e3EBz5j GlotLID : Language identification for low-resource languages . In The 2023 Conference on Empirical Methods in Natural Language Processing

  10. [18]

    Amr Keleg and Walid Magdy. 2023. https://doi.org/10.18653/v1/2023.arabicnlp-1.31 A rabic dialect identification under scrutiny: Limitations of single-label classification . In Proceedings of ArabicNLP 2023, pages 385--398, Singapore (Hybrid). Association for Computational Linguistics

  11. [19]

    Bård Borch Michalsen. 2015. Komma. Kommategnets personlighet, historie og regler. Juritzen forlag

  12. [20]

    Aleksandra Mileti \'c and Filip Mileti \'c . 2024. https://aclanthology.org/2024.humeval-1.4 A gold standard with silver linings: Scaling up annotation for distinguishing B osnian, C roatian, M ontenegrin and S erbian . In Proceedings of the Fourth Workshop on Human Evaluation...

  13. [21]

    Manning, Ryan McDonald, Slav Petrov, Sampo Pyysalo, Natalia Silveira, Reut Tsarfaty, and Daniel Zeman

    Joakim Nivre, Marie-Catherine de Marneffe, Filip Ginter, Yoav Goldberg, Jan Haji c , Christopher D. Manning, Ryan McDonald, Slav Petrov, Sampo Pyysalo, Natalia Silveira, Reut Tsarfaty, and Daniel Zeman. 2016. https://aclanthology.org/L16-1262 U niversal D ependencies v1: A mul...

  14. [22]

    Manning, Sampo Pyysalo, Sebastian Schuster, Francis Tyers, and Daniel Zeman

    Joakim Nivre, Marie-Catherine de Marneffe, Filip Ginter, Jan Haji c , Christopher D. Manning, Sampo Pyysalo, Sebastian Schuster, Francis Tyers, and Daniel Zeman. 2020. https://aclanthology.org/2020.lrec-1.497 U niversal D ependencies v2: An evergrowing multilingual treebank co...

  15. [23]

    NLLB Team , Marta R. Costa-jussà, James Cross, Onur Çelebi, Maha Elbayad, Kenneth Heafield, Kevin Heffernan, Elahe Kalbassi, Janice Lam, Daniel Licht, Jean Maillard, Anna Sun, Skyler Wang, Guillaume Wenzek, Al Youngblood, Bapi Akula, Loic Barrault, Gabriel Mejia Gonzalez, Pran...

  16. [24]

    Eric Papazian. 2013. Moltke moe og norsk språknormering fram til 1907. Språklig samling, pages 69--104

  17. [25]

    Alec Radford, Jeff Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. 2019. Language models are unsupervised multitask learners

  18. [26]

    Javier de la Rosa and Per Egil Kummervold. 2022. https://huggingface.co/NbAiLab/nb-nordic-lid/tree/main NB - N ordic- LID

  19. [27]

    David Samuel, Andrey Kutuzov, Samia Touileb, Erik Velldal, Lilja vrelid, Egil R nningstad, Elina Sigdel, and Anna Palatkina. 2023. https://aclanthology.org/2023.nodalida-1.61 N or B ench -- a benchmark for N orwegian language models . In Proceedings of the 24th Nordic Conferen...

  20. [28]

    David Samuel, Vladislav Mikhailov, Erik Velldal, Lilja Øvrelid, Lucas Georges Gabriel Charpentier, and Andrey Kutuzov. 2024. http://arxiv.org/abs/2412.06484 Small languages, big models: A study of continual training on languages of norway

  21. [29]

    Victor Sanh, Lysandre Debut, Julien Chaumond, and Thomas Wolf. 2019. https://arxiv.org/abs/1910.01108 D istil BERT , a distilled version of BERT : smaller, faster, cheaper and lighter . ArXiv, abs/1910.01108

  22. [30]

    Raivis Skadi n s , J \"o rg Tiedemann, Roberts Rozis, and Daiga Deksne. 2014. http://www.lrec-conf.org/proceedings/lrec2014/pdf/846_Paper.pdf Billions of parallel words for free: Building and using the EU bookshop corpus . In Proceedings of the Ninth International Conference o...

  23. [31]

    V \'e steinn Sn bjarnarson, Annika Simonsen, Goran Glava s , and Ivan Vuli \'c . 2023. https://aclanthology.org/2023.nodalida-1.74 Transfer to a low-resource language via close relatives: The case study on F aroese . In Proceedings of the 24th Nordic Conference on Computationa...

  24. [32]

    J \"o rg Tiedemann. 2020. https://aclanthology.org/2020.wmt-1.139 The tatoeba translation challenge -- realistic data sets for low resource and multilingual MT . In Proceedings of the Fifth Conference on Machine Translation, pages 1174--1182, Online. Association for Computatio...

  25. [33]

    Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pierric Cistac, Tim Rault, Remi Louf, Morgan Funtowicz, Joe Davison, Sam Shleifer, Patrick von Platen, Clara Ma, Yacine Jernite, Julien Plu, Canwen Xu, Teven Le Scao, Sylvain Gugger, Mari...

  26. [34]

    Marcos Zampieri, Kai North, Tommi Jauhiainen, Mariano Felice, Neha Kumari, Nishant Nair, and Yash Mahesh Bangera. 2024. https://aclanthology.org/2024.lrec-main.882 Language variety identification with true labels . In Proceedings of the 2024 Joint International Conference on C...

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.