---
title: "Automatic Difficulty Classification of Arabic Sentences (2021)"
description: "Khallaf and Sharoff build what they present as the first sentence-level difficulty classifier for Modern… 40 references, cited by 9 papers here."
authors: ["Nouran Khallaf", "Serge Sharoff"]
year: 2021
venue: "Proceedings of the Sixth Arabic Natural Language Processing Workshop"
source: "https://aclanthology.org/2021.wanlp-1.11/"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/khallaf-2021-automatic-difficulty-classification-arabic-sentences/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/khallaf-2021-automatic-difficulty-classification-arabic-sentences/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2021](https://md.research.elamur.ai/free/en/papers/2021/page/6/) / Automatic Difficulty Classification of Arabic Sentences

# Automatic Difficulty Classification of Arabic Sentences

**Authors:** [Nouran Khallaf](https://md.research.elamur.ai/free/en/authors/nouran-khallaf/) and [Serge Sharoff](https://md.research.elamur.ai/free/en/authors/serge-sharoff/) · **Year:** 2021 · **Venue:** [Proceedings of the Sixth Arabic Natural Language Processing Workshop](https://md.research.elamur.ai/free/en/venues/wanlp/) ([WANLP 2021](https://md.research.elamur.ai/free/en/venues/wanlp/2021/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2021.wanlp-1.11/) · [PDF](https://aclanthology.org/2021.wanlp-1.11.pdf) · [OpenAlex](https://openalex.org/W3155149882)

**Citations:** 7 citations (OpenAlex, 2026-10-14) · [9](https://md.research.elamur.ai/free/en/papers/khallaf-2021-automatic-difficulty-classification-arabic-sentences/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
Arabic readability work has mostly scored whole documents, but automatic text simplification needs to know which individual sentences are too hard. Khallaf and Sharoff build what they present as the first sentence-level difficulty classifier for Modern Standard Arabic, aimed at learners of Arabic as a foreign language. They create training data by giving each sentence the document-level label of its learner-oriented source (Gloss, ALC, the Al-Kitaab textbook). Long sentences from the I-AR web corpus serve as advanced-level material. Noisy labels are cleaned by flagging sentences where an ensemble of classifiers unanimously contradicts the gold label. The authors compare hand-built features with fastText, mBERT, XLM-R and Arabic-BERT representations. The features are POS from MADAMIRA, dependency structure from CamelParser, and a new CEFR-tagged lemma list. The task is framed both as classification and as regression. Fine-tuned Arabic-BERT performs best, reaching F-1 0.80 on three CEFR bands and 0.94 on simple vs complex. The feature-based models are weaker but transfer better to an unseen corpus and make predictions easier to interpret. The result is a practical tool for picking suitable texts for learners.

## Key points
- Dataset One has about 22.7k sentences labelled A/B/C. The A and B sentences come from Gloss (ILR levels mapped to CEFR), the ALC learner corpus and Al-Kitaab (ACTFL levels mapped to CEFR). The C level uses 8,627 I-AR web sentences longer than 15 words.
- For label cleaning, five classifiers vote: SVM (rbf), Random Forest, KNeighbors, Softmax and XgBoost. In 1,943 sentences all five disagreed with the gold label. An Arabic teacher reviewed a sample and tagged cases as Wrong, Modify, Ambiguous or False. In the end, 380 sentences were moved to a lower level, mostly from B to A.
- The feature set has 34 features: POS and morphology, dependency-based syntax, and CEFR lexical incidence plus word entropy. The lexical features rely on a merged list of 8,834 lemmas built from Buckwalter and Parkinson, KELLY and the Al-Kitaab chapter lists. Sentence length was deliberately left out.
- Reported results:
  - 3-way classification: F-1 0.80 for Arabic-BERT and 0.75 for XLM-R.
  - Regression: Spearman 0.71.
  - Binary classification: F-1 0.94.
- Dataset Two contains 2,980 original/simplified sentence pairs from the novel "Saaq al-Bambuu" and its learner edition. A pair-similarity classifier on this data (AraBert, Arabic-BERT) reaches F-1 0.98.

## Interesting ideas
- When the gold labels are suspect, let several models from different ML families vote, and send only the cases where they all overrule the gold label to a human expert with a small error taxonomy; this concentrates scarce expert time on the most likely label errors.
- The authors drop sentence length as a feature because length correlates with difficulty without causing it, which keeps the model from learning a cheap shortcut.
- Each word's difficulty is turned into a distribution over CEFR bands, and the entropy of that distribution across the sentence becomes a single lexical-difficulty feature.
- In the pair-similarity task, the original sentence copied unchanged is used as a negative example next to its true simplification, so the model must learn whether a change happened, not just that the two texts are related.
- Interpretable feature models were weaker in-domain but transferred better across corpora than BERT, so it pays to keep both and test them on a different evaluation corpus.

## For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

- **I1** (learn): The witness-parser silver pipeline could route trees where all witnesses unanimously contradict the treebank to review under Wrong/Modify/Ambiguous/False tags, separating parser errors from treebank errors.
- **I2** (answer): A difficulty or complexity estimate for test questions should be built from structural signals such as clause count and connectors, not raw length, so that its explanations stay honest.
- **I3** (world): WordNet/frequency knowledge could carry a CEFR-like band per lemma, letting Mova explain which words make a text hard for a given reader.
- **I4** (back): Evaluation of the text → graph → text compressor could include identity pairs as negative controls, to check that the simplify/paraphrase detector does not reward a verbatim copy.
- **I5** (learn): Evaluation of the explainable parser against the GPU parser should report out-of-domain transfer separately, since the explainable model may be the more robust choice there.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

🧭 Topics (1): [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) · 2,908 works

## Cite

Cite as: Nouran Khallaf and Serge Sharoff. 2021. Automatic Difficulty Classification of Arabic Sentences. Proceedings of the Sixth Arabic Natural Language Processing Workshop.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{khallaf2021automatic,
  title = {{Automatic Difficulty Classification of Arabic Sentences}},
  author = {Nouran Khallaf and Serge Sharoff},
  booktitle = {Proceedings of the Sixth Arabic Natural Language Processing Workshop},
  year = {2021},
  url = {https://aclanthology.org/2021.wanlp-1.11/}
}
```

- [← previous AraCOVID19-MFH: Arabic COVID-19 Multi-label Fake News and Hate Speech Detection Dataset](https://md.research.elamur.ai/free/en/papers/ameur-2021-aracovid19-mfh-arabic-covid-19/)
- [next → Coordinate Constructions in English Enhanced Universal Dependencies: Analysis and Computational Modeling](https://md.research.elamur.ai/free/en/papers/grunewald-2021-coordinate-constructions-english-enhanced-universal/)
- [↑ back to papers of 2021](https://md.research.elamur.ai/free/en/papers/2021/page/6/)

## References

40 references, 11 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Hend S Al-Khalifa and Amani A Al-Ajlan. 2010. Automatic readability measurements of the arabic text: An exploratory study. Arabian Journal for Science and Engineering, 35(2 C):103–124. [find](https://scholar.google.com/scholar?q=Hend+S+Al%2DKhalifa+and+Amani+A+Al%2DAjlan%2E+2010%2E+Automatic+readability+measurements+of+the+arabic+text%3A+An+exploratory+study%2E+Arabian+Journal+for+Science+and+Engineering%2C+35%282+C%29%3A103%E2%80%93124%2E)
2. Saud Al-Sanousi. 2013. Saaq al-Bambuu. Arab Scientific Publishers Inc., Lebanon. [find](https://scholar.google.com/scholar?q=Saud+Al%2DSanousi%2E+2013%2E+Saaq+al%2DBambuu%2E+Arab+Scientific+Publishers+Inc%2E%2C+Lebanon%2E)
3. Abdel Karim Al Tamimi, Manar Jaradat, Nuha Al-Jarrah, and Sahar Ghanem. 2014. Aari: automatic arabic readability index. Int. Arab J. Inf. Technol., 11(4):370–378. [find](https://scholar.google.com/scholar?q=Abdel+Karim+Al+Tamimi%2C+Manar+Jaradat%2C+Nuha+Al%2DJarrah%2C+and+Sahar+Ghanem%2E+2014%2E+Aari%3A+automatic+arabic+readability+index%2E+Int%2E+Arab+J%2E+Inf%2E+Technol%2E%2C+11%284%29%3A370%E2%80%93378%2E)
4. A. Alfaifi and Eric S Atwell. 2013. Arabic learner corpus v1: A new resource for arabic language research. In In proceedings of the Second Workshop on Arabic Corpus Linguistics (WACL-2). Leeds. [find](https://scholar.google.com/scholar?q=A%2E+Alfaifi+and+Eric+S+Atwell%2E+2013%2E+Arabic+learner+corpus+v1%3A+A+new+resource+for+arabic+language+research%2E+In+In+proceedings+of+the+Second+Workshop+on+Arabic+Corpus+Linguistics+%28WACL%2D2%29%2E+Leeds%2E)
5. Amal Alsaif. 2012. Human and automatic annotation of discourse relations for Arabic. University of Leeds. [find](https://scholar.google.com/scholar?q=Amal+Alsaif%2E+2012%2E+Human+and+automatic+annotation+of+discourse+relations+for+Arabic%2E+University+of+Leeds%2E)
6. Wissam Antoun, Fady Baly, and Hazem Hajj. 2020. [Arabert: Transformer-based model for arabic language understanding](https://md.research.elamur.ai/free/en/papers/antoun-2020-arabert-transformer-based-model-arabic/). arXiv preprint arXiv:2003.00104. [↗](https://arxiv.org/abs/2003.00104)
7. K Brustad, M Al-Batal, and A Al-Tonsi. 2015. Al-Kitaab fii Tacallum al-cArabiyya. A Textbook for Beginning Arabic: Part One Third Edition. Georgetown University Press. [find](https://scholar.google.com/scholar?q=K+Brustad%2C+M+Al%2DBatal%2C+and+A+Al%2DTonsi%2E+2015%2E+Al%2DKitaab+fii+Tacallum+al%2DcArabiyya%2E+A+Textbook+for+Beginning+Arabic%3A+Part+One+Third+Edition%2E+Georgetown+University+Press%2E)
8. Tim Buckwalter and Dilworth Parkinson. 2014. A frequency dictionary of Arabic: Core vocabulary for learners. Routledge. [↗](https://doi.org/10.4324/9780203883280)
9. Violetta Cavalli-Sforza, Mariam El Mezouar, and Hind Saddiki. 2014. Matching an arabic text to a learners’ curriculum. In Proc. 5th Int. Conf. on Arabic Language Processing (CITALA), Oujda, Morocco, pages 79–88. [find](https://scholar.google.com/scholar?q=Violetta+Cavalli%2DSforza%2C+Mariam+El+Mezouar%2C+and+Hind+Saddiki%2E+2014%2E+Matching+an+arabic+text+to+a+learners%E2%80%99+curriculum%2E+In+Proc%2E+5th+Int%2E+Conf%2E+on+Arabic+Language+Processing+%28CITALA%29%2C+Oujda%2C+Morocco%2C+p)
10. Violetta Cavalli-Sforza, Hind Saddiki, and Naoual Nassiri. 2018. Arabic readability research: Current state and future directions. Procedia computer science, 142:38–49. [find](https://scholar.google.com/scholar?q=Violetta+Cavalli%2DSforza%2C+Hind+Saddiki%2C+and+Naoual+Nassiri%2E+2018%2E+Arabic+readability+research%3A+Current+state+and+future+directions%2E+Procedia+computer+science%2C+142%3A38%E2%80%9349%2E)
11. Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer, and Veselin Stoyanov. 2019. Unsupervised cross-lingual representation learning at scale. arXiv preprint arXiv:1911.02116. [↗](https://arxiv.org/abs/1911.02116)
12. Nuraihan Mat Daud, Haslina Hassan, and Nor-maziah Abdul Aziz. 2013. A corpus-based readability formula for estimate of arabic texts reading difficulty. World Applied Sciences Journal, 21:168–173. [find](https://scholar.google.com/scholar?q=Nuraihan+Mat+Daud%2C+Haslina+Hassan%2C+and+Nor%2Dmaziah+Abdul+Aziz%2E+2013%2E+A+corpus%2Dbased+readability+formula+for+estimate+of+arabic+texts+reading+difficulty%2E+World+Applied+Sciences+Journal%2C+21%3A168%E2%80%93173%2E)
13. Felice Dell’Orletta, Simonetta Montemagni, and Giulia Venturi. 2014. Assessing document and sentence readability in less resourced languages and across textual genres. ITL-International Journal of Applied Linguistics, 165(2):163–193. [↗](https://doi.org/10.1075/itl.165.2.03del)
14. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. [Bert: Pre-training of deep bidirectional transformers for language understanding](https://md.research.elamur.ai/free/en/papers/devlin-2019-bert-pre-training-deep-bidirectional/). arXiv preprint arXiv:1810.04805. [↗](https://arxiv.org/abs/1810.04805)
15. Marilena Di Bari, Serge Sharoff, and Martin Thomas. 2014. Multiple views as aid to linguistic annotation error analysis. In Proceedings of LAW VIII-The 8th Linguistic Annotation Workshop, pages 82–86. [find](https://scholar.google.com/scholar?q=Marilena+Di+Bari%2C+Serge+Sharoff%2C+and+Martin+Thomas%2E+2014%2E+Multiple+views+as+aid+to+linguistic+annotation+error+analysis%2E+In+Proceedings+of+LAW+VIII%2DThe+8th+Linguistic+Annotation+Workshop%2C+pages+82%E2%80%9386%2E)
16. William H DuBay. 2004. The principles of readability. Online Submission. [find](https://scholar.google.com/scholar?q=William+H+DuBay%2E+2004%2E+The+principles+of+readability%2E+Online+Submission%2E)
17. Mahmoud El-Haj and Paul Rayson. 2016. Osman—a novel arabic readability metric. In Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC’16), pages 250–255. [find](https://scholar.google.com/scholar?q=Mahmoud+El%2DHaj+and+Paul+Rayson%2E+2016%2E+Osman%E2%80%94a+novel+arabic+readability+metric%2E+In+Proceedings+of+the+Tenth+International+Conference+on+Language+Resources+and+Evaluation+%28LREC%E2%80%9916%29%2C+pages+250%E2%80%93255%2E)
18. Laila Familiar. 2016. Saud al-Sanousi’s Saaq al-Bambuu: The Authorized Abridged Edition for Students of Arabic. Georgetown University Press. [find](https://scholar.google.com/scholar?q=Laila+Familiar%2E+2016%2E+Saud+al%2DSanousi%E2%80%99s+Saaq+al%2DBambuu%3A+The+Authorized+Abridged+Edition+for+Students+of+Arabic%2E+Georgetown+University+Press%2E)
19. Jonathan Neil Forsyth. 2014. Automatic readability prediction for modern standard Arabic. Ph.D. thesis, Brigham Young University. Department of Linguistics and English Language. [find](https://scholar.google.com/scholar?q=Jonathan+Neil+Forsyth%2E+2014%2E+Automatic+readability+prediction+for+modern+standard+Arabic%2E+Ph%2ED%2E+thesis%2C+Brigham+Young+University%2E+Department+of+Linguistics+and+English+Language%2E)
20. Edouard Grave, Piotr Bojanowski, Prakhar Gupta, Armand Joulin, and Tomas Mikolov. 2018. [Learning word vectors for 157 languages](https://md.research.elamur.ai/free/en/papers/grave-2018-learning-word-vectors-157-languages/). Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018) [Online].
21. Adam Kilgarriff, Frieda Charalabopoulou, Maria Gavrilidou, Janne Bondi Johannessen, Saussan Khalil, Sofie Johansson Kokkinakis, Robert Lew, Serge Sharoff, Ravikiran Vadlapudi, and Elena Volodina. 2014. Corpus-based vocabulary lists for language learners for nine languages. Language resources and evaluation, 48(1):121–163. [↗](https://doi.org/10.1007/s10579-013-9251-2)
22. Gerry Knowles and Zuraidah Mohd Don. 2004. The notion of a “lemma”: Headwords, roots and lexical sets. International Journal of Corpus Linguistics, 9(1):69–81. [↗](https://doi.org/10.1075/ijcl.9.1.04kno)
23. Kendall Maurice and Gibbons Jean Dickinson. 1990. Rank correlation methods. London: Edward Arnold. [find](https://scholar.google.com/scholar?q=Kendall+Maurice+and+Gibbons+Jean+Dickinson%2E+1990%2E+Rank+correlation+methods%2E+London%3A+Edward+Arnold%2E)
24. Naoual Nassiri, Abdelhak Lakhouaja, and Violetta Cavalli-Sforza. 2018a. Arabic readability assessment for foreign language learners. In International Conference on Applications of Natural Language to Information Systems, pages 480–488. Springer. [find](https://scholar.google.com/scholar?q=Naoual+Nassiri%2C+Abdelhak+Lakhouaja%2C+and+Violetta+Cavalli%2DSforza%2E+2018a%2E+Arabic+readability+assessment+for+foreign+language+learners%2E+In+International+Conference+on+Applications+of+Natural+Language+to+)
25. Naoual Nassiri, Abdelhak Lakhouaja, and Violetta Cavalli-Sforza. 2018b. Modern standard arabic readability prediction. In International Conference on Arabic Language Processing, pages 120–133. Springer. [find](https://scholar.google.com/scholar?q=Naoual+Nassiri%2C+Abdelhak+Lakhouaja%2C+and+Violetta+Cavalli%2DSforza%2E+2018b%2E+Modern+standard+arabic+readability+prediction%2E+In+International+Conference+on+Arabic+Language+Processing%2C+pages+120%E2%80%93133%2E+Springe)
26. Arfath Pasha, Mohamed Al-Badrashiny, Mona T Diab, Ahmed El Kholy, Ramy Eskander, Nizar Habash, Manoj Pooleery, Owen Rambow, and Ryan Roth. 2014. [Madamira: A fast, comprehensive tool for morphological analysis and disambiguation of arabic](https://md.research.elamur.ai/free/en/papers/pasha-2014-madamira-fast-comprehensive-tool-morphological/). In Lrec, volume 14, pages 1094–1101.
27. Anna Rogers, Olga Kovaleva, and Anna Rumshisky. 2020. [A primer in BERTology: What we know about how BERT works](https://md.research.elamur.ai/free/en/papers/rogers-2020-primer-bertology-what-we-know/). Transactions of the Association for Computational Linguistics, 8:842–866. [↗](https://doi.org/10.1162/tacl_a_00349)
28. Hind Saddiki, Karim Bouzoubaa, and Violetta Cavalli-Sforza. 2015. Text readability for arabic as a foreign language. In 2015 IEEE/ACS 12th International Conference of Computer Systems and Applications (AICCSA), pages 1–8. IEEE. [find](https://scholar.google.com/scholar?q=Hind+Saddiki%2C+Karim+Bouzoubaa%2C+and+Violetta+Cavalli%2DSforza%2E+2015%2E+Text+readability+for+arabic+as+a+foreign+language%2E+In+2015+IEEE%2FACS+12th+International+Conference+of+Computer+Systems+and+Applications)
29. Hind Saddiki, Nizar Habash, Violetta Cavalli-Sforza, and Muhamed Al Khalil. 2018. [Feature optimization for predicting readability of arabic l1 and l2](https://md.research.elamur.ai/free/en/papers/saddiki-2018-feature-optimization-predicting-readability-arabic/). In Proceedings of the 5th Workshop on Natural Language Processing Techniques for Educational Applications, pages 20–29. [↗](https://doi.org/10.60692/v8hb7-91m29)
30. Ali Safaya, Moutasem Abdullatif, and Deniz Yuret. 2020. [Kuisail at semeval-2020 task 12: Bert-cnn for offensive speech identification in social media](https://md.research.elamur.ai/free/en/papers/safaya-2020-kuisail-semeval-2020-task-12/). In Proceedings of the Fourteenth Workshop on Semantic Evaluation, pages 2054–2059.
31. Horacio Saggion. 2017. Automatic text simplification. Synthesis Lectures on Human Language Technologies, 10(1):1–137. [↗](https://doi.org/10.1007/978-3-031-02166-4)
32. Elliot Schumacher, Maxine Eskenazi, Gwen Frishkoff, and Kevyn Collins-Thompson. 2016. Predicting the relative difficulty of single sentences with and without surrounding context. arXiv preprint arXiv:1606.08425. [↗](https://arxiv.org/abs/1606.08425)
33. Sarah Schwarm and Mari Ostendorf. 2005. [Reading level assessment using support vector machines and statistical language models](https://md.research.elamur.ai/free/en/papers/schwarm-2005-reading-level-assessment-support-vector/). In Proceedings of the 43rd Annual Meeting of the Association for Computational Linguistics (ACL’05), pages 523–530, Ann Arbor, Michigan. Association for Computational Linguistics.
34. Anas Shahrour, Salam Khalifa, Dima Taji, and Nizar Habash. 2016. [Camelparser: A system for arabic syntactic analysis and morphological disambiguation](https://md.research.elamur.ai/free/en/papers/shahrour-2016-camelparser-system-arabic-syntactic-analysis/). In Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: System Demonstrations, pages 228–232.
35. Serge Sharoff. 2006. Open-source corpora: using the net to fish for linguistic data. International Journal of Corpus Linguistics, 11(4):435–462. [find](https://scholar.google.com/scholar?q=Serge+Sharoff%2E+2006%2E+Open%2Dsource+corpora%3A+using+the+net+to+fish+for+linguistic+data%2E+International+Journal+of+Corpus+Linguistics%2C+11%284%29%3A435%E2%80%93462%2E)
36. Serge Sharoff. 2021. Genre annotation for the web: text-external and text-internal perspectives. Register studies, 3. [↗](https://doi.org/10.1075/rs.19015.sha)
37. E Tschirner, O Bärenfänger, and K Wisniewski. 2015. Assessing evidence of validity of the actfl cefr listening and reading proficiency tests (lpt and rpt) using a standard-setting approach. Technical Report 2015-EU-PUB-2. [find](https://scholar.google.com/scholar?q=E+Tschirner%2C+O+Ba%CC%88renfa%CC%88nger%2C+and+K+Wisniewski%2E+2015%2E+Assessing+evidence+of+validity+of+the+actfl+cefr+listening+and+reading+proficiency+tests+%28lpt+and+rpt%29+using+a+standard%2Dsetting+approach%2E+Technica)
38. Sowmya Vajjala and Kaidi Loo. 2014. Automatic cefr level prediction for estonian learner text. In Proceedings of the third workshop on NLP for computer-assisted language learning, pages 113–127. [find](https://scholar.google.com/scholar?q=Sowmya+Vajjala+and+Kaidi+Loo%2E+2014%2E+Automatic+cefr+level+prediction+for+estonian+learner+text%2E+In+Proceedings+of+the+third+workshop+on+NLP+for+computer%2Dassisted+language+learning%2C+pages+113%E2%80%93127%2E)
39. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, ukasz Kaiser, and Illia Polosukhin. 2017. [Attention is all you need](https://md.research.elamur.ai/free/en/papers/vaswani-2017-attention-all-need/). In Advances in neural information processing systems, pages 5998–6008.
40. Yinfei Yang, Daniel Cer, Amin Ahmad, Mandy Guo, Jax Law, Noah Constant, Gustavo Hernandez Abrego, Steve Yuan, Chris Tar, Yun-Hsuan Sung, et al. 2019. [Multilingual universal sentence encoder for semantic retrieval](https://md.research.elamur.ai/free/en/papers/yang-2020-multilingual-universal-sentence-encoder-semantic/). arXiv preprint arXiv:1907.04307. [↗](https://arxiv.org/abs/1907.04307)

*Updated 1 November 2026*